BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
MEETING ABSTRACTS
24th Annual Computational Neuroscience
Meeting: CNS*2015
Prague, Czech Republic. 18-23 July 2015
Edited by Gennady Cymbalyuk and Anthony Burkitt
Published: 4 December 2015
These abstracts are available online at http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
KEYNOTE PRESENTATIONS
A1
Modeling cortical dynamics with Wilson-Cowan equations
Jack Cowan
Mathematics Department, Neurology Department, and Committee on
Computational Neuroscience, University of Chicago, Chicago, IL, USA
E-mail: cowan@math.uchicago.edu
BMC Neuroscience 2015, 16(Suppl 1):A1
Experimental data collected over the last decade indicates that there exist
at least two distinct modes of cortical response to stimuli. In mode 1 a
low intensity stimulus triggers a wave that propagates at a velocity of
about 0.3 m/sec, with an amplitude that decays exponentially. In mode 2
a high intensity stimulus triggers a larger response that remains local,
and does not propagate to neighboring regions. Other data indicate that
unstimulated or resting cortex exhibits pair correlations between
neighboring cells, the amplitudes of which decay slowly with distance,
whereas stimulated cortex exhibits pair correlations whose amplitude falls
of rapidly with distance. Here we show how the mean-field Wilson-Cowan
equations can account precisely for the two modes of cortical response,
and how stochastic Wilson-Cowan equations can account for the
behavior of the pair correlations. We will present these results after
outlining the basic properties of both the mean-field and stochastic
equations.
A2
The dynamics of resting fluctuations in the brain
Gustavo Deco
Center for Brain and Cognition, Universitat Pompeu Fabra / ICREA, Barcelona,
Spain
E-mail: gustavo.deco@upf.edu
BMC Neuroscience 2015, 16(Suppl 1):A2
The grand average functional connectivity (FC) of a resting brain captures
properly the well- structured spatial correlations between different brain
areas. Whole-brain-models explicitly linking spontaneous local neuronal
dynamics with the tractography based anatomical structure of the brain
are able to explain the emergence of those spatial resting correlations.
Nevertheless, resting activity is not only spatially structured but also
shows a very stereotypical temporal structure which is characterized by
rapid transitions switching between a few discrete FC states across time.
In this talk, we introduce a powerful theoretical framework, which allows
us to demonstrate that resting functional connectivity FC dynamics (FCD)
constrains more strongly the dynamical working point of whole-brain
models. Furthermore, using a very general neural mass model based on
2015 various authors All articles published in this supplement are distributed under the terms of the Creative Commons Attribution
License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium,
provided the original work is properly cited.
Open Access
the normal form of a Hopf bifurcation we are able to demonstrate that the
temporal dynamics of resting state fluctuations emerges at the edge of
the transition between asynchronous to oscillatory behavior. Even more
importantly, at that particular working point the global metastability of the
whole brain is maximized. By optimizing the spectral characteristics of
each local brain node, we discover the dynamical core of the brain, i.e., the
set of nodes, which drives the rest of the whole brain by oscillations.
FEATURED ORAL PRESENTATIONS
F1
Complex synapses as efficient memory systems
Marcus K Benna*, Stefano Fusi
Center for Theoretical Neuroscience, Columbia University, College of
Physicians and Surgeons, New York, NY 10032, USA
E-mail: mkb2162@columbia.edu
BMC Neuroscience 2015, 16(Suppl 1):F1
The molecular machinery underlying memory consolidation at the level of
synaptic connections is believed to employ a complex network of highly
diverse biochemical processes that operate on a wide range of different
timescales. An appropriate theoretical framework could help us identify
their computational roles and understand how these intricate networks of
interactions support synaptic memory formation and maintenance.
Here we construct a broad class of synaptic models that can efficiently
harness biological complexity to store and preserve a huge number of
memories, vastly outperforming other synaptic models of memory. The
number of storable memories grows almost linearly with the number of
synapses, which constitutes a substantial improvement over the square
root scaling of previous models [1,2], especially when large neural systems
are considered. This improvement is obtained without significantly
reducing the initial memory strength, which still scales approximately like
the square root of the number of synapses.
This is achieved by combining together multiple dynamical processes that
operate on different timescales, to ensure the memory strength decays as
slowly as the inverse square root of the age of the corresponding synaptic
modification. Memories are initially stored in fast variables and then
progressively transferred to slower ones. Importantly, in our case the
interactions between fast and slow variables are bidirectional, in contrast
to the unidirectional cascades of previous models.
The proposed models are robust to perturbations of parameters and can
capture several properties of biological memories, which include delayed
expression of synaptic potentiation and depression, synaptic metaplasticity,
and spacing effects. We discuss predictions for the autocorrelation function
of the synaptic efficacy that can be tested in plasticity experiments involving
long sequences of synaptic modifications.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
References
1. Fusi S, Drew PJ, Abbott LF: Cascade models of synaptically stored
memories. Neuron 2005, 45(4):599-611.
2. Roxin A, Fusi S: Efficient partitioning of memory systems and its importance
for memory consolidation. PLoS Comput Biol 2013, 9(7):e1003146.
F2
Closed-loop approach to tuning deep brain stimulation parameters for
Parkinsons disease
Abbey B Holt1*, Max Shinn2, Theoden I Netoff1,3
1
Graduate Program in Neuroscience, University of Minnesota, Minneapolis,
MN, 55455, USA; 2Department of Neuroscience, University of Minnesota,
Minneapolis, MN, 55455, USA; 3Department of Biomedical Engineering,
University of Minnesota, Minneapolis, MN, 55455, USA
E-mail: holt0437@umn.edu
BMC Neuroscience 2015, 16(Suppl 1):F2
Deep brain stimulation (DBS) is used to treat motor symptoms of patients
with Parkinsons disease (PD). However, tuning stimulation parameters is
currently done using a time intensive trial-and-error process until
maximum therapy is achieved with minimal side effects [1]. There is a
need for a systematic approach to tuning parameters based on patient
physiology. With the development of DBS electrodes that can
simultaneously stimulate and record [2], a closed-loop approach may be
taken. It is hypothesized that emergent oscillations in the basal ganglia
network, particularly in the beta range (12-35 Hz) lead to motor symptoms
of PD [3], and that DBS works by disrupting these oscillations. Our
hypothesis is that stimulating at a specific phase in the pathological
oscillation will optimally disrupt the oscillatory activity, and that this phase
can be predicted from the phase response curve (PRC). Here, we use a
computational network model of PD with an emergent pathological 34 Hz
oscillation [4] to test this closed-loop approach to DBS and confirm the
results in vitro. By stimulating at a specific phase in the beta oscillation we
are able to modulate the power of the oscillation in the model. By
stimulating soon after the peak in the oscillation, we disrupt the 34 Hz
oscillation, while stimulating later in the period enhances it. Hence, the
timing of stimulation affects how well the population of neurons
desynchronized. Next, we test this concept in vitro by synchronizing patch-
clamped neurons in the substantia nigra pars reticulata (an output nucleus
of the basal ganglia) to an oscillatory input, such as a beta oscillation. We
show that stimulating at a particular phase of the oscillatory input affects
how well neurons synchronize or desynchronize to that input. Finally, we
show it is possible to use the PRC to predict how stimulating at a specific
phase will affect the neurons ability to synchronize or desynchronize from
the oscillatory input in vitro.
This work shows that stimulating at specific phases in an oscillation can
synchronize or desynchronize neurons in a computational model and in
vitro. By stimulating at specific phases of an emergent pathological
oscillation in a closed-loop approach to DBS, we were able to suppress a
pathological oscillation in a computational model of PD. In this approach, a
frequency of 34 Hz was used for DBS, which is much lower than the value
used clinically (>100 Hz). Through closed-loop stimulation, precisely timed
stimuli with respect to the phase of the oscillation can dramatically decrease
stimulus power needed for DBS. The ability to synchronize or desynchronize
a neuron to an oscillatory input by stimulating at a certain phase was also
validated in vitro. It is possible to predict the phase of stimulation to
maximally disrupt neuronal synchronization to an external oscillatory input
in single neurons using a PRC. We have previously shown a novel method
to estimate a PRC from population data [5] in a computational model of PD.
This suggests it may be possible to predict the phase at which to stimulate
in order to optimally disrupt a pathological population oscillation in PD
using the PRC, and apply this in a closed-loop approach to DBS.
Acknowledgements: Research supported by MnDrive Neuromodulation
Fellowship, NSF Collaborative Research Grant, and Neuroengineering NSF
IGERT under DGE-1069104
References
1. Volkmann J, Herzog J, Kopper F, Deuschl G: Introduction to the
programming of deep brain stimulators. Movement Disorders 2002,
17(Suppl 3):S181-S187.
2. Ryapolova-Webb E, Afshar P, Stanslaski S, Denison T, de Hemptinne C,
Bankiewicz K, Starr PA: Chronic cortical and electromyographic recordings
Page 2 of 200
from a fully implantable device: preclinical experience in a nonhuman
primate. J Neural Eng 2014, 11(1):016009.
3. Dostrovsky J, Bergman H: Oscillatory activity in the basal ganglia
relationship to normal physiology and pathophysiology. Brain 2004,
127(Pt 4):721-722.
4. Hahn PJ, McIntyre CC: Modeling shifts in the rate and pattern of
subthalamopallidal network activity during deep brain stimulation.
Journal of Computational Neuroscience 2010, 28(3):425-441.
5. Holt AB, Netoff TI: Origins and suppression of oscillations in a
computational model of Parkinsons disease. J Comput Neurosci 2014,
37(3):505-521.
F3
Control of gamma vs beta competition in olfactory bulb by the balance
between sensory input and centrifugal feedback control
Franois David1, Emmanuelle Courtiol1,2, Nathalie Buonviso1,
Nicolas Fourcaud-Trocm1*
1
Lyon Neuroscience Research Center, CNRS UMR 5292, INSERM U1028,
Universit Claude Bernard, Lyon, France; 2Emotional Brain Institute, Nathan
Kline Institute for Psychiatric Research and the Department of Child and
Adolescent Psychiatry New York University Langone Medical Center, New
York, NY, USA
E-mail: nicolas.fourcaud-trocme@cnrs.fr
BMC Neuroscience 2015, 16(Suppl 1):F3
Gamma (40-80Hz) and beta (15-40Hz) oscillations and their associated
neuronal assemblies are key features of neuronal sensory processing.
However, the mechanisms involved in either their interaction and/or the
switch between these different regimes in most sensory systems remain
misunderstood. The mammalian olfactory bulb (OB) expresses both
gamma and beta oscillations, which appear to be mutually exclusive, and a
slower one related to respiration (2-10Hz). Gamma oscillations have been
linked to odorant physical properties (quality, intensity) while beta
oscillations are strongly increased by odor experience (for reviews see
[1,2]). Importantly, the occurrence pattern of these two fast alternating
oscillations is intermingled with the respiratory slow rhythm which
provides a window for odor discrimination. Based on in vivo recordings
and biophysical modeling of the mammalian olfactory bulb (OB), we
explored how OB internal dynamics and the balance between sensory and
centrifugal inputs control the occurrence and alternation of OB gamma
and beta oscillations over a respiratory cycle.
In the OB, fast oscillations originate in the dendrodendritic interaction
between excitatory mitral cells (MCs) and inhibitory granule cells (GCs).
Experimental evidence has shown that GC dendritic arbor can operate in
two modes: a local mode which effectively allows a weak inhibition
between MCs without requiring GC spikes, and a global mode which
induces a strong inhibition of MCs following GC spikes. We implemented
these two inhibitory mechanisms in a parsimonious and flexible OB model
based on generalized integrate-and-fire models. In the granule non-spiking
regime, the weak inhibition can sustain OB oscillation in the gamma
frequency range with characteristics of an auto-entrainment process [3]. In
contrast, in the granular spiking regime, MCs sufficiently excite the GCs
such that the latter discharge and induce a strong inhibitory input which
silences the MC population and generate beta oscillations, similarly to the
PING regime [4]. Intrinsic properties of each type of oscillation are
remarkably stable regarding most of tested network parameters. However
their occurrence depends strongly on OB network sensory and centrifugal
inputs (onto MCs and GCs respectively). In particular, sensory activation of
MCs must be strong enough for the emergence of gamma oscillations,
while sufficient centrifugal activation of GCs, to allow them to spike, is
necessary to generate beta. Based on novel experimental data in
anesthetized rat, we show that both inputs are slowly modulated by the
respiratory rhythm but phase shifted by about a quarter cycle. In our
model, this phase shift can account for the gamma-beta alternation
observed in vivo. Finally, additional tests show that the model captures
accurately the competition between gamma and beta oscillations when
sensory or centrifugal inputs are modulated such as in different natural
conditions involving odor characteristics (odor intensity) and behavior
(odor experience, active sniffing).
Overall the model approaches very closely OB dynamics observed in vivo,
and can thus be used to interpret present and future experiments.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
References
1. Martin C, Ravel N: Beta and gamma oscillatory activities associated with
olfactory memory tasks: different rhythms for different functional
networks? Frontiers in Behavioral Neuroscience 2014, 8:218.
2. Kay LM: Circuit oscillations in odor perception and memory. Prog in Brain
Research 2014, 208:223-251.
3. David FO, Hugues E, Cenier T, Fourcaud-Trocm N, Buonviso N: Specific
Entrainment of Mitral Cells during Gamma Oscillation in the Rat
Olfactory Bulb. PLoS Computational Biology 2009, 5(10):19.
4. Brgers C, Kopell N: Synchronization in networks of excitatory and
inhibitory neurons with sparse, random connectivity. Neural Computation
2003, 15(3):509-538.
ORAL PRESENTATIONS
O1
Limits to the scalability of cortical network models
Sacha J van Albada1*, Moritz Helias1, Markus Diesmann1,2,3
1
Institute of Neuroscience and Medicine (INM-6) and Institute for Advanced
Simulation (IAS-6), Jlich Research Centre and JARA, Jlich, Germany;
2
Department of Psychiatry, Psychotherapy and Psychosomatics, Medical
Faculty, RWTH Aachen University, Aachen, Germany; 3Department of Physics,
Faculty 1, RWTH Aachen University, Aachen, Germany
E-mail: s.van.albada@fz-juelich.de
BMC Neuroscience 2015, 16(Suppl 1):O1
The size of the mammalian brain is inconveniently right in the middle
between a few interacting particles and a mole of matter on a logarithmic
scale. In physics, we learn that often in the limit where system size goes to
infinity, simple mathematical expressions can be obtained uncovering the
mechanisms governing the dynamics at the large but finite system size in
nature. In neuroscience, however, we found that such an ansatz may fail
because correlations drop so slowly that the mechanism governing the
behavior in the infinite size limit [1] is not the mechanism relevant at the
scale of the brain circuit in question [2]. The direct simulation of networks
at their natural size has historically been difficult due to the sheer number
of neurons and synapses. Therefore, neuroscientists also routinely explore
the other side of the logarithmic scale and investigate downscaled circuits.
In summary, it seems that brain networks are often too small for the
infinity limit and too large for simulations.
In this contribution, we assess the scalability of networks in the asynchronous
irregular state with a focus on downscaling. By extending the theory of
correlations in such networks [2-5] and verifying analytical predictions by
direct simulations using NEST [6], we formally demonstrate that generally
already second-order measures cannot be preserved. The underlying
mathematical reason is a one-to-one mapping between correlation structure
and effective connectivity, which depends both on the physical connectivity
and on the working point of the neurons [7]. Correlations are relevant
because they influence synaptic plasticity [8] and large-scale measurements
of neuronal activity [9], and are related to information processing and
behavior [10,11].
Our results show that the reducibility of asynchronous networks is
fundamentally limited, indicating the importance of considering networks
with realistic numbers of neurons and synapses. Fortunately, corresponding
simulation technology is becoming available to neuroscience [12]. Both the
investigation of the infinity limit and the exploration of downscaled
networks remain powerful methods of computational neuroscience.
However, researchers should make explicit the rationale they apply in up- or
downscaling.
Acknowledgements: We acknowledge funding by the Helmholtz
Association: portfolio theme SMHB and Young Investigators Group VH-
NG-1028, and EU Grants 269921 (BrainScaleS) and 604102 (Human Brain
Project)
References
1. Renart A, de la Rocha J, Bartho P, Hollender L, Parga N, Reyes A, Harris KD:
The asynchronous state in cortical circuits. Science 2010,
327(5965):587-590.
2. Helias M, Tetzlaff T, Diesmann M: The correlation structure of local cortical
networks intrinsically results from recurrent dynamics. PLoS Comput Biol
2014, 10(1):e1003428.
Page 3 of 200
3. Ginzburg I, Sompolinsky H: Theory of correlations in stochastic neural
networks. Phys Rev E 1994, 50(4):3171-3191.
4. Grytskyy D, Tetzlaff T, Diesmann M, Helias M: A unified view on weakly
correlated recurrent networks. Front Comput Neurosci 2013, 7:131.
5. Helias M, Tetzlaff T, Diesmann M: Echoes in correlated neural systems.
New J Phys 2013, 15:023002.
6. Gewaltig MO, Diesmann M: NEST (NEural Simulation Tool). Scholarpedia
2007, 2:1430.
7. van Albada SJ, Helias M, Diesmann M: Scalability of asynchronous
networks is limited by one-to-one mapping between effective
connectivity and correlations. arXiv preprint 2014, 1411:4770.
8. Morrison A, Aertsen A, Diesmann M: Spike-timing dependent plasticity in
balanced random networks. Neural Comput 2007, 19(6):1437-1467.
9. Lindn H, Tetzlaff T, Potjans TC, Pettersen KH, Grn S, et al: Modeling the
spatial reach of the LFP. Neuron 2011, 72(5):859-872.
10. Zohary E, Shadlen MN, Newsome WT: Correlated neuronal discharge rate
and its implications for psychophysical performance. Nature 1994,
370(6485):140-143.
11. Riehle A, Grn S, Diesmann M, Aertsen A: Spike synchronization and rate
modulation differentially involved in motor cortical function. Science
1997, 278(5345):1950-1953.
12. Kunkel S, Schmidt M, Eppler JM, Plesser HE, Masumoto G, et al: Spiking
network simulation code for petascale computers. Front Neuroinform
2014, 8:78.
O2
The high-conductance state enables neural sampling in networks of
LIF neurons
Mihai A Petrovici1*, Ilja Bytschok1, Johannes Bill2, Johannes Schemmel1,
Karlheinz Meier1
1
Kirchhoff-Institute for Physics, University of Heidelberg, Heidelberg,
Germany; 2Institute for Theoretical Computer Science, University of Graz,
Graz, Austria
E-mail: mpedro@kip.uni-heidelberg.de
BMC Neuroscience 2015, 16(Suppl 1):O2
The apparent stochasticity of in-vivo neural circuits has long been
hypothesized to represent a signature of ongoing stochastic inference in
the brain [1-3]. More recently, a theoretical framework for neural sampling
has been proposed, which explains how sample-based inference can be
performed by networks of spiking neurons [4,5]. One particular
requirement of this approach is that the membrane potential of these
neurons satisfies the so-called neural computability condition (NCC), which
in turn leads to a logistic neural response function.
Analytical approaches to calculating this function have been the subject of
many theoretical studies. In order to make the problem tractable, particular
assumptions regarding the neural or synaptic parameters are usually made
[6,7]. However, biologically significant activity regimes exist which are not
covered by these approaches: Under strong synaptic bombardment, as is
often the case in cortex, the neuron is shifted into a high-conductance state
(HCS), which is characterized by a small membrane time constant. In this
regime, synaptic time constants and refractory periods dominate membrane
dynamics.
The HCS is also particularly interesting from a functional point of view. In
[5], we have shown that LIF neurons that are shifted into a HCS by
background synaptic bombardment can attain the correct firing statistics
to sample from well-defined probability distributions (i.e., satisfy the NCC).
In order to calculate the response function of neurons in this regime, we
are required to consider a new approach.
The core idea of this approach is to separately consider two different
modes of spiking dynamics: burst spiking and transient quiescence, in
which the neuron does not spike for longer periods. For the bursting
mode, we explicitly take into consideration the autocorrelation of the
membrane potential before and after refractoriness by propagating the
PDF of the effective membrane potential from spike to spike within a
burst. For the membrane potential evolution between bursts, we consider
an Ornstein-Uhlenbeck approximation. We find that our theoretical
prediction of the neural response function closely matches simulation data.
Moreover, in the HCS scenario, we show that the neural response function
becomes symmetric and can be well approximated by a logistic function,
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 4 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

thereby providing the correct dynamics in order to perform neural

sampling. Such stochastic firing units can then be used to sample from
arbitrary probability distributions over binary random variables [4,5,8,9].
We hereby provide not only a normative framework for Bayesian inference
in cortex, but also powerful applications of low-power, accelerated
neuromorphic systems to highly relevant machine learning problems.
Acknowledgements: This research was supported by EU grants #269921
(BrainScaleS), #237955 (FACETS-ITN), #604102 (Human Brain Project), the
Austrian Science Fund FWF #I753-N23 (PNEUMA) and the Manfred Strk
Foundation
References
1. Krding K, Wolpert D: Bayesian integration in sensorimotor learning.
Nature 2004, 427:244-247.
2. Fizser J, Berkes P, Orbn G, Lengyel M: Statistically optimal perception and
learning: from behavior to neural representations. Trends in Cognitive
Sciences 2010, 14(3):119-130.
3. Friston K, Mattout J, Kilner J: Action understanding and active inference.
Biological Cybernetics 2011, 104(1-2):137-160.
4. Bsing Lars, Bill Johannes, Nessler Bernhard, Maass Wolfgang: Neural
dynamics as sampling: A model for stochastic computation in recurrent
networks of spiking neurons. PLoS Computational Biology 2011, 7(11):
e1002211.
5. Petrovici MA, Bill J, Bytschok I, Schemmel J, Karlheinz Meier: Stochastic
inference with deterministic spiking neurons. arXiv preprint 2013,
1311.3211.
6. Brunel N, Sergi S: Firing Frequency of leaky integrate-and-fire neurons
with synaptic current dynamics. Journal of Theoretical Biology 1998,
195(1):87-95.
7. Moreno-Bote R, Parga N: Role of synaptic filtering on the firing response
of simple model neurons. Physical Review Letters 2004, 92:028102.
8. Pecevski D, Bsing L, Maass W: Probabilistic inference in general graphical
models through sampling in stochastic networks of spiking neurons.
PLoS Computational Biology 2011, 7(12):e1002294.
9. Probst D, Petrovici MA, Bytschok I, Bill J, Peceyski D, Schemmel J, Meier K:
Probabilistic inference in discrete spaces can be implemented into
networks of LIF neurons. Frontiers in Neuroscience 2015, 9:13.

Figure 1(abstract O3) Estimated branching ratio s in dependence

O3
Quantifying the distance to criticality under subsampling
Jens Wilting1*, Viola Priesemann1,2
1
Max-Planck-Institute for Dynamics and Self-Organization, D-37077
Gttingen, Germany; 2Bernstein Center for Computational Neuroscience,
University of Gttingen, D-37075 Gttingen, Germany
E-mail: jwilting@nld.ds.mpg.de
BMC Neuroscience 2015, 16(Suppl 1):O3
Neuronal systems have been proposed to operate close to criticality. But
how far from criticality are they precisely? We developed a novel method
to determine the distance to criticality from data. Importantly, our method
is reliable under subsampling, i.e. the experimental constraint that in many
dynamical systems only a small fraction of all agents can be sampled.
Thereby, our novel approach for the first time allows to determine the
distance to criticality without bias from spiking activity in vivo, which in
general is strongly subsampled.
In more detail, neuronal systems have been proposed to operate close to
criticality, as power-law distributions of the avalanche size have been
found for local field potentials from in vitro preparations [1], to human
cortex [2]. Criticality is an attractive candidate state for neural dynamics,
because in models criticality maximizes processing capacities [3]. However,
it has been widely overlooked that criticality also comes with the risk of
spontaneous runaway activity (epilepsy). Recent experiments suggest that
spiking activity in rats, cats, and monkeys, is in a sub-critical regime,
keeping a safety-margin from criticality [4]. Quantifying the precise
distance to criticality may help to shed light on how the brain maximizes
its information processing capacities without risking runaway activity.
In neural systems, critical dynamics is usually studied in the context of
branching processes with continuous drive [1], because they approximate
well the functional propagation of spiking activity on the network [4]. The
dynamics of branching processes are determined by the expected
number of spikes s in postsynaptic neurons triggered by a single spike,
showing either stationary dynamics (sub-critical, s < 1) or transient
of sampled units n of a system of size N, for conventional (empty
symbols) and our novel (full) measures in theory and models and
in spike recordings
growth (super-critical, s > 1); for s = 1 branching processes are critical
and generate the characteristic power law scaling. Methods to infer s
from fully sampled systems are well established, however, subsampling
[5] resulted in strongly biased estimates (Fig 1., empty symbols). To
overcome this bias, we derived a novel measure, based on a multistep
linear regression. This measure for the first time allows to quantify the
distance to criticality even under strong subsampling (Fig., full symbols).
Our method generalizes to auto-regressive processes with both additive
and multiplicative noise, making it widely applicable. We validated our
method by applying subsampling to simulated branching processes with
invasion, and to a generic integrate-and-fire model. After validation, we
applied this method to highly parallel spike recordings from macaque
prefrontal cortex, cat visual cortex, and rat hippocampus. These analysis
indicated that spiking activity is clearly subcritical (0.97 < s < 0.99; N =
10 experiments), and not critical.
References
1. Beggs J, Plenz D: Neuronal avalanches in neocortical circuits. J Neurosci
2003, 23(35):11167-11177.
2. Priesemann V, Valerrame M, Wibral M, Le Van Quyen M: Neuronal
avalanches differ from wakefulness to deep sleep - evidence from
intracranial depth recordings in humans. PLoS Comp Biol 2013, 9(3):
e1002985.
3. Boedecker J, Obst O, Lizier JT, Mayer NM, Asada M: Information processing
in echo state networks at the edge of chaos. Theory Biosci 2012,
131(3):205-213.
4. Priesemann V, Wibral M, Valderrama M, Prpper R, Le Van Quyen M,
Geisel T, Triesch J, Nikolic D, Munk MHJ: Spike avalanches in vivo suggest
a driven, slightly subcritical brain state. Front Syst Neurosci 2014, 8:108.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
5. Priesemann V, Munk MHJ, Wibral M: Subsampling effects in neuronal
avalanche distributions recorded in vivo. BMC Neuroscience 2009, 10:40.
O4
Large-scale analysis of brain-wide electrophysiological diversity reveals
novel characterization of mammalian neuron types
Shreejoy J Tripathy*, Dmitry Tebaykin, Brenna Li, Ogan Marcarci, Lilah Toker,
Paul Pavlidis
Centre for High-Throughput Biology, University of British Columbia,
Vancouver, BC V6K 2B7, Canada
E-mail: stripat3@gmail.com
BMC Neuroscience 2015, 16(Suppl 1):O4
Brains achieve efficient function through implementing a division of labor,
in which different neurons serve distinct computational roles. One striking
Figure 1(abstract O4) Hierarchical clustering of diverse neuron types on the basis of electrophysiological similarity. Electrophysiological
parameters are obtained from the NeuroElectro database via literature-mining and are normalized to account for variability in experimental
methodologies across studies
Page 5 of 200
way in which neuron types differ is in their electrophysiology properties.
These properties arise through expression of combinations of ion channels
that collectively define the computations that a neuron performs on its
inputs and its role within its larger circuit. Though the electrophysiology of
many neuron types has been previously characterized, these data exist
across thousands of journal articles, making cross-study neuron-to-neuron
comparisons difficult.
Here, we present NeuroElectro, a public database where physiological
properties for the majority of mammalian neuron types have been
compiled through semi-automated literature text-mining and expert
curation. The corresponding web application, at http://www.neuroelectro.
org, provides a rich dynamic interface for visualizing and comparing
physiological information across neuron types; conveniently linking
extracted data back to its primary reference. Mining the database content
after normalization for methodological differences, we show that there
exist but 5-9 major neuron classes in terms of electrophysiological
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 6 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

properties, which separate largely based on cell size and basal levels of
excitability (Figure 1).
As an example of how this resource can help answer fundamental
questions in neuroscience, we integrate NeuroElectro with neuronal gene
expression from public datasets like the Allen Brain Atlas. We show that
simple statistical models can accurately predict features of a neurons
electrophysiological phenotype given information of its gene expression
alone. We further investigate these models to ask which genes, of the 20K
in the genome, are most predictive of neuron physiology. We find that
while ion channel-related genes provide significant predictive power, the
most predictive gene classes surprisingly correspond to G-proteins and
transcription factors, suggesting the involvement of hundreds of diverse
genes in regulating a neurons computational function.
O5
Self-organization of computation in neural systems by interaction
between homeostatic and synaptic plasticity
Sakyasingha Dasgupta1,2*, Christian Tetzlaff1,2*, Tomas Kulvicius3,
Florentin Wrgtter1,2
1
Institute for Physics - Biophysics, Georg-August-University, D-37077
Gttingen, Germany; 2Bernstein Center for Computational Neuroscience,
Georg-August-University, D-37077 Gttingen, Germany; 3Maersk-Moller
Mckinsey Institute, Southern Denmark University, Odense, Denmark
E-mail: sdasgup@phys.uni-goettingen.de; tetzlaff@phys.uni-goettingen.de
BMC Neuroscience 2015, 16(Suppl 1):O5
The ability to perform complex motor control tasks is essentially enabled
by the nervous system via the self-organization of large groups of neurons
into coherent dynamic activity patterns. During learning, this is brought
about by synaptic plasticity, resulting in the formation of multiple
functional networks - commonly termed cell-assemblies. A multitude of
such cell assemblies provide the requisite machinery for non-linear
computations needed for the mastery of a large number of motor skills.
However, given the fact that there exists considerable overlap between the
usage of the same neurons within such assemblies, for a wide range of
motor tasks, creation and sustenance of such computationally powerful
networks poses a challenging problem. How such interwoven assembly
networks self-organize and how powerful assemblies can coexist therein,
without catastrophically interfering with each other remains largely
unknown. One the one side, it is already known that networks can be
trained to perform complex nonlinear calculations [1], such that, if the
network possesses a reservoir of rich, transient dynamics, desired outputs
can be extracted from these reservoirs in order to enable motor control.
On the other side, cell assemblies are created by Hebbian learning rules

Figure 1(abstract O5) Cell assembly size and computational performance are correlated. (A) Input-driven formation of cell assemblies brought
about by the interaction long-term potentiation (LTP) and synaptic scaling (Syn. Sca.). (B) With more learning trials the assembly grows and integrates
more neurons. We measure this by arbitrarily defining assembly size by that set of neurons connected with efficacies larger than half the maximum
weights. (C) Parallel to the outgrowth of the cell assembly the error of the system to perform several linear and non-linear calculations decreases
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
that strengthen a synapse if pre- and post-synaptic neurons are co-active
within a small enough time window [2]. Therefore it appears relatively
straightforward to combine these mechanisms in order to construct
powerful assembly networks. However, given that the self-organization of
neurons into cell assemblies by the processes of synaptic plasticity induces
ordered or synchronized neuronal dynamics, which can destroy the
required complexity of a reservoir network, such a combination remains a
very challenging problem [3]. Furthermore, simultaneous creation of
multiple cell assemblies can also lead to catastrophic interference if one
cannot prevent them from growing into each other. In this study, we
exploit for the first time the interaction between neuronal and synaptic
processes acting on different time scales to enable, on a long time scale,
the self-organized formation of assembly networks (Fig. 1), while on a
shorter timescale, to conjointly perform several non-linear calculations
needed for motor fine-control. Specifically, by the combination of synaptic
plasticity and synaptic scaling [4], as a homeostatic mechanism, we
demonstrate that such self-organization allows executing a difficult, six
degrees of freedom, manipulation task with a robot where assemblies
need to learn computing complex nonlinear transforms and - for execution
- must cooperate with each other without interference. This mechanism,
thus, permits for the first time, the guided self-organization of
computationally powerful sub-structures in dynamic networks for behavior
control. Furthermore, comparing our assembly network to networks with
unchanging synapses ("static networks) shows that it is indeed the
embedding of a strongly connected assembly that creates the necessary
computational power.
References
1. Buonomano DV, Maass W: State-dependent computations:
spatiotemporal processing in cortical networks. Nat. Rev Neurosci 2009,
10:113-125.
2. Palm G, Knoblauch A, Hauser F, Schultz A: Cell assemblies in the cerebral
cortex. Biol Cybern 2014, 108:559-572.
3. Klamp S, Maass W: Emergence of dynamic memory traces in cortical
microcircuit models through STDP. J Neurosci 2013,
33(28):11515-11529.
4. Tetzlaff C, Kolodziejski C, Timme M, Tsodyks M, Wrgtter F: Synaptic
scaling enables dynamically distinct short- and long-term memory
formation. PLoS Comput Biol 2013, 9(10):e10003307.
O6
A model for spatially periodic firing in the hippocampal formation
based on interacting excitatory and inhibitory plasticity
Simon N Weber1,2*, Henning Sprekeler1,2
1
Technische Universitt Berlin, 10587, Berlin, Germany; 2Bernstein Center for
Computational Neuroscience, 10115, Berlin, Germany
E-mail: weber@tu-berlin.de
BMC Neuroscience 2015, 16(Suppl 1):O6
Figure 1(abstract O6) Example for the emergence of a grid cell. Columns from left to right: Spatial tuning of excitatory and inhibitory inputs (two
examples each); spatial activity pattern of the output neuron before and after learning; auto-correlogram of activity after learning
Page 7 of 200
Neurons in the hippocampal formation exhibit a variety of spatially tuned
firing patterns. The mechanisms by which these different patterns emerge
are not fully resolved, although competing computational models exist for
several of them. Here we present a new model that can generate all
observed spatial firing patterns by a single mechanism. The model consists
of a feedforward network with a single output neuron. Its essential
ingredients are i) spatially tuned excitatory and inhibitory inputs [e.g., 1]
and ii) interacting excitatory and inhibitory Hebbian plasticity. The
inhibitory plasticity homeostatically controls the output firing rate by
balancing excitation and inhibition [2]. We show in simulations and by a
mathematical analysis that the output neuron develops periodic firing
patterns along a stimulus dimension if inhibitory inputs are more broadly
tuned than excitatory inputs along this dimension. More generally,
depending on the relative spatial auto-correlation length of the excitatory
and inhibitory inputs, the model exhibits firing patterns that are similar to
those of place cells, grid cells (see Figure 1) or band cells (neurons that fire
on spatially periodic bands [3]). For inputs with combined spatial and head
direction tuning, the same mechanism leads to output firing patterns
reminiscent of head direction cells and conjunctive cells (neurons that fire
like grid cells in space but only at a particular head direction). A linear
stability analysis of the homogeneous steady state accurately predicts the
spatial periodicity obtained from simulations. The model combines the
robust pattern formation of attractor models [e.g., 4], with the spatial
(rather than neural) structure formation of models based on synaptic
plasticity [5]. In contrast to attractor models [6], our model predicts that
the grid spacing should be robust to global modifications in inhibitory
synaptic strength, a distinction which could be experimentally verified.
In conclusion, we propose a feedforward network model that generates
all known spatial firing patterns in the hippocampal formation through a
single self-organizing mechanism.
Acknowledgements: Funded by the German Federal Ministry for
Education and Research, FKZ 01GQ1201
References
1. Buetfering C, Allen K, Monyer H: Parvalbumin interneurons provide grid
cell-driven recurrent inhibition in the medial entorhinal cortex. Nature
Neurosci 2014, 17(5):710-718.
2. Vogels TP, Sprekeler H, Zenke F, Clopath C, Gerstner W: Inhibitory plasticity
balances excitation and inhibition in sensory pathways and memory
networks. Science 2011, 334(6062):1569-1573.
3. Krupic J, Burgess N, OKeefe J: Neural representations of location
composed of spatially periodic bands. Science 2012, 337(6096):853-857.
4. Fuhs MC, Touretzky DS: A spin glass model of path integration in rat
medial entorhinal cortex. J Neurosci 2006, 26(16):4266-4276.
5. Kropff E, Treves A: The emergence of grid cells: Intelligent design or just
adaptation? Hippocampus 2008, 18(12):1256-1269.
6. Couey JJ, Witoelar A, Zhang SJ, Zheng K, Ye J, Dunn B, et al: Recurrent
inhibitory circuitry as a mechanism for grid formation. Nature Neurosci
2013, 16(3):318-324.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
O7
Whole-cell morphological properties of neurons constrain the
nonrandom features of network connectivity
Jugoslava Aimovi1*, Tuomo Mki-Marttunen2, Marja-Leena Linne1
1
Department of Signal Processing, Tampere University of Technology,
Tampere, Finland; 2Institute of Clinical Medicine, University of Oslo, Oslo,
Norway
E-mail: jugoslava.acimovic@tut.fi
BMC Neuroscience 2015, 16(Suppl 1):O7
We addressed the principles of micro-level organization of neuronal
circuits and explored how the neuronal morphology constrains this
organization. Several studies have demonstrated the non-trivial properties
of the network connectivity using in vitro recordings from multiple
neurons [1-3], yet it is unclear to what extent this structure reflects
reorganization caused by synaptic plasticity, and what is imposed by the
morphological constraints. Two recent articles explored this issue using the
simulated neural circuits and demonstrated the specific structural
properties in those circuits [4,5].
We analyzed a model that emphasizes the role of single-cell morphology,
a homogeneous population of neurons in a planar space without
boundaries. Each neuron is composed of two displaced neurite fields
defined on the limited support. A neurite field describes the likelihood of
finding a neurite segment at a certain point in the plane. Using a
proximity criterion (Peters rule) the expected number of potential
synapses is estimated between each pair of neurons. Alternatively, this
number can be estimated from the realistic morphology of a simulated
neuron, or from the morphologies reconstructed from in vitro/in vivo
recordings. The number of potential synapses depends on the axon-
dendrite distance, which leads to a definition of the expected radius. An
axon-dendrite pair that is expected to form at least one synapse must be
on a distance not larger than the effective radius. All considered
statistical measures of network connectivity are expressed as the
functions of the effective radius normalized with the neuron size. In this
study, we considered the standard graph theoretic measures of network
connectivity, the motif counts, clustering coefficient, path length, and
small-world coefficient. It has been demonstrated that they have a
significant impact on the population activity in simulated networks [6].
Changing the normalized effective radius from small (<0.3) to big (>10) we
vary the network properties between the two extremes. For the small
values of the effective radius, the networks favor unidirectional
connections and sparse local connectivity. The clustering coefficient and
the path length are similar to those obtained in uniform random networks,
i.e. in the networks independent of topology. For the large values of the
effective radius, the local connectivity is dense with the majority of
bidirectional connections. As the normalized effective radius increases, the
clustering coefficient increases towards the values obtained for the
networks with dominant local connectivity, while the path length remains
close to the one of the uniform random networks. The normalized
effective radius on the interval 1-2, provides the biggest variability of
connectivity patterns and the optimized properties relevant for the
information transfer.
Conclusions: We present a theoretical framework that relates
neuromorphology with the connectivity in neuronal circuits, and that can be
solved analytically. The normalized effective radius was found to be the key
morphological property that dominantly affects considered connectivity
measures. By tuning it we can obtain the networks with the biggest
variability of local connectivity patterns. At the same time, those networks
acquire the key characteristics of the small-world networks, known to
optimize the information transfer.
References
1. Song S, Sjstrm PJ, Reigl M, Nelson S, Chklovskii DB: Highly nonrandom
features of synaptic connectivity in local cortical circuits. PLoS Biol 2005,
3(3):e68.
2. Rieubland S, Roth A, Husser M: Structured connectivity in cerebellar
inhibitory networks. Neuron 2014, 81(4):913-929.
3. Perin R, Berger T, Markram H: A synaptic organizing principle for cortical
neuronal groups. Proc Natl Acad Sci 2011, 108:5419-5424.
4. Perin R, Telefont M, Markram H: Computing the size and number of
neuronal clusters in local circuits. Front Neuroanat 2013, 7:1.
Page 8 of 200
5. McAssey MP, Bijma F, Tarigan B, van Pelt J, van Ooyen A, de Gunst M: A
morpho density approach to estimating neural connectivity. PLoS One
2014, 9(1):e86526.
6. Mki-Marttunen T, Aimovi J, Ruohonen K, Linne M-L: Structure-Dynamics
Relationships in Bursting Neuronal Networks Revealed Using a
Prediction Framework. PLoS One 2013, 8(7):e69373.
O8
Origin of the kink of somatic action potentials
Maria T Teleczuk*, Marcel Stimberg, Romain Brette
University Pierre et Marie Curie, Paris, 75006, France
E-mail: maria@telenczuk.pl
BMC Neuroscience 2015, 16(Suppl 1):O8
The Hodgkin and Huxley (1952) model of action potential (AP) generation
accounts for many properties of APs observed experimentally and has been
successfully used in modeling neurons of different types. In this model,
however, the spike onset is much shallower than in experimental recordings
from the soma suggesting different activation properties of sodium channels
in the real tissue. To explain the origin of the observed sharpness (kink) in the
spike onset three hypotheses were proposed: 1. Cooperative hypothesis:
sodium channels cooperate in the axon initial segment, which makes their
collective activation curve much sharper [1]. However, there is no
experimental evidence for this hypothesis. 2. Active backpropagation
hypothesis: spikes are initiated in the axon and backpropagate to the soma.
The kink is caused by the sharpening of the axonal spike by active
conductances during its backpropagation through the axon [2]. 3.
Compartmentalization hypothesis: the kink comes from distal initiation and
the current sink caused by the difference in the size of the soma and axon [3].
To find out what is truly happening in the cell during the action potential,
we investigated the active backpropagation and compartmentalization
hypotheses by means of computational modeling and theoretical analysis.
In order to differentiate the hypotheses, we varied systematically the
morphology of the neuron and distribution of the ionic channels along the
cell, and tested how they contribute to the appearance of the kink. We
show that the kink at spike onset is primarily due to compartmentalization
rather than to active backpropagation.
Acknowledgements: MTT is the recipient of an cole des Neurosciences
de Paris (ENP) Graduate Program fellowship and the Rgion Ile-de-France
(DIM Cerveau et Pense). This work was also supported by Agence
Nationale de la Recherche (ANR-14-CE13-0003)
References
1. Naundorf B, Wolf F, Volgushev M: Unique features of action potential
initiation in cortical neurons. Nature 2006, 440(7087):1060-1063.
2. Yu Y, Shu Y, McCormick D: Cortical action potential backpropagation
explains spike threshold variability and rapid-onset kinetics. J Neurosci
2008, 28(29):7260-7272.
3. Brette R: Sharpness of spike initiation in neurons explained by
compartmentalization. PLoS Comp Biol 2013, 9(12):e1003338.
Figure 1(abstract O8) Kink in the action potential. Patch clamp
recordings (red) from a cortical pyramidal cell and action potential
produced by a Hodgkin Huxley type model (black). Left: Voltage-time
relationship. Right: Phase plot of the same traces as in the left (dV/dt vs. V).
Note, in both representations the onset of the action potential is much
faster for the experimental recordings (red) than for the model (black)
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
O9
An accurate circuit-based description of retinal ganglion cell
computation
Yuwei Cui1, Yanbin V Wang2, Jonathan B Demb2, Daniel A Butts
1
Department of Biology and Program in Neuroscience and Cognitive
Science, University of Maryland, College Park, MD 20742, USA; 2Department
of Ophthalmology and Visual Science and Department of Cellular and
Molecular Physiology, Yale University, New Haven, CT 06511 USA
E-mail: dab@umd.edu
BMC Neuroscience 2015, 16(Suppl 1):O9
Visual processing depends on computations performed by complex neural
circuits. Although the circuitry in the retina has been extensively
characterized, common functional models of how ganglion cell spike trains
represent visual stimuli typically rely on linear descriptions of their receptive
field [1]. Different types of nonlinear models have offered improvements in
spike train prediction, but such improvements are often incremental, and in
most cases not linked to known elements of the retinal circuit.
Here, we describe a new nonlinear model framework designed to
represent key elements of the retinal circuit, which can predict recorded
retinal ganglion cell spike trains with high temporal precision. We used
recordings of both synaptic currents (via voltage-clamp recordings) and
spike (via loose patch recordings) from the same ON Alpha ganglion cells
in the mouse retina in order to build a two-stage nonlinear model. This
model describes ganglion cell computation as sums and products of
excitatory and inhibitory inputs [2]. Model parameters were estimated
based on either intracellular or spike train data using a maximum-
likelihood framework.
We found that excitatory synaptic currents to the ganglion cell are well
described by an excitatory input combined with divisive suppression,
both elements described by LN models fit to intracellular data. Using
stimuli with center-surround structure, we demonstrate that this divisive
suppression arises from the surround, and is the likely result of
presynaptic inhibition mediated by amacrine cells [3], rather than
synaptic depression [4]. We then extended this nonlinear model of
synaptic currents to explain spike response of the ganglion cell by
incorporating a spiking nonlinearity with spike refractoriness. All model
parameters could be fit using the spike trains alone, resulting in a
prediction of the excitatory currents that closely matched the models fit
directly to the currents.
The resulting model had unprecedented ability to predict both synaptic
current and spike trains (with >90% of the explainable variance) at one
millisecond resolution on cross-validation datasets, capturing both fast
transient responses in synaptic current, as well as the high precision of
spike train responses. Furthermore, the model output automatically
adapted to contrast, and could predict the responses across contrast
levels with similar accuracy without any change in model parameters.
Notably, the nonlinear structure of the model was particular to ON Alpha
ganglion cells, and other retinal ganglion cell types had distinct
computational structures, likely corresponding to different underlying
connectivity within the retina governing their processing of vision.
Thus, by targeting a nonlinear model based on the specific computations
performed by retinal circuit elements, we uncovered an extremely
accurate description of retinal processing, and identified two-stage
computational properties that can be linked to elements of the retinal
circuit. In addition to providing an accurate description of ON Alpha cells,
such computational framework also sets a foundation for understanding
the different roles of the ~20 ganglion cell types that comprise the input
to the rest of the visual system.
References
1. Chichilnisky EJ: A simple white noise analysis of neuronal light responses.
Network 2001, 12:199-213.
2. Euler T, Haverkamp S, Schubert T: Retinal bipolar cells: elementary
building blocks of vision. Nature Rev Neurosci 2014, 15(8):507-519.
3. McFarland JM, Cui Y, Butts DA: Inferring nonlinear neuronal computation
based on physiologically plausible inputs. PLoS Computational Biology
2013, 9(7):e1003143.
4. Ozuysal Y, Baccus SA: Linking the computational structure of variance
adaptation to biophysical mechanisms. Neuron 2012, 73(5):1002-1015.
Page 9 of 200
O10
Contrast-dependent modulation of gamma rhythm in v1: a network
model
Margarita Zachariou1*, Mark Roberts2,3, Eric Lowet2,3, Peter de Weerd2,3,
Avgis Hadjipapas1,4
1
University of Nicosia Medical School, Nicosia, Cyprus; 2Psychology and
Neuroscience, Maastricht University, Maastricht, the Netherlands; 3Donders
Institute for Brain, Cognition and Behaviour, Radboud University Nijmegen,
Nijmegen, the Netherlands; 4St. Georges University of London, Cranmer
Terrace, London, SW17 0RE, UK
E-mail: zachariou.m@unic.ac.cy
BMC Neuroscience 2015, 16(Suppl 1):O10
In our empirical data comprising of single-unit and LFP recordings in
macaque area V1 and source reconstructed human MEG localized to
visual cortex we have observed a robust increase in gamma oscillation
frequency with increasing luminance contrast. In addition, at high grating
contrasts, a robust decay in gamma power was observed in the LFP [1]
but not the MEG. These phenomena are key to understanding the
functional role of network frequencies and for investigating the stability
of gamma oscillations at both local and macroscopic levels. However,
even at the most basic level of spatially- undifferentiated neuronal
models, it is not fully understood how excitatory (E) and inhibitory (I)
neurons interact to generate the observed network gamma oscillations in
the macaque single-unit and LFP data. For example we could obtain the
frequency shift and power decay in a network where the rhythm is
produced by excitatory neurons that fired more frequently than inhibitory
neurons, and in another more neurophysiologically plausible network
composed of excitatory neurons showing sparse firing [2,3] and inhibitory
neurons showing faster firing [4]. Moreover, it is unknown how increasing
excitatory afferent drive (of which luminance contrast is a proxy)
modulates the interactions between E and I populations (as well as
interactions within each population) to account for changes in frequency
and power. We aimed to replicate the empirical data from macaque
visual cortex and to further investigate the stability of the observed
gamma oscillation. Here, we present an undifferentiated V1 network PING
model, with realistic neuronal features as determined and validated from
the analysis of a large number of V1 neurons obtained in 3 rhesus
monkeys. The model when perturbed by increasing afferent input,
exhibits the core characteristics of the empirical data, that is, (1) a
monotonic increase in LFP frequency, (2) a non-monotonic LFP power
modulation with decay at high inputs, (3) a largely non-saturating
increase in average unit firing rate. In addition, the model exhibits
realistic single unit behavior across a range of inputs. In terms of the
frequency shift, we have observed remarkable scaling behaviour: while
the frequency of oscillations changes dramatically with input, the
absolute average phase at which inhibitory and excitatory neurons fire in
each oscillation cycle and the average relative phase to each other
remain constant. This scaling may on one hand underlie the stability of
the gamma oscillation locally and on other hand facilitate communication
through coherence in the gamma range [5] across varying stimulus
conditions, by preserving the timing and relative ordering of population
firing irrespective of the oscillation frequency [6]. Our results suggest that
the observed power decline results from a primary (functional)
decoupling among inhibitory neurons. Further analysis highlighted that
the functional decoupling is related to the balance of inhibition/
excitation. In further steps, we intend to test these predictions in the
empirical data, and then proceed to a differentiated V1 columnar model
to investigate the divergences between human MEG and macaque LFP/
spiking responses.
Acknowledgements: Research was supported by the European Union
Seventh Framework Programme (FP7/2007-2013) under grant agreement
no. 604102 (Human Brain Project) to P.D.W. and A.H
References
1. Roberts MJ, Lowet E, Brunet NM, Wal Ter M, Tiesinga P, Fries P, De Weerd P:
Robust gamma coherence between macaque V1 and V2 by dynamic
frequency matching. Neuron 2013, 78(3):523-536.
2. Wang XJ: Neurophysiological and computational principles of cortical
rhythms in cognition. Physiol Rev 2010, 90(3):1195-1268.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
3. Kilpatrick ZP, Ermentrout B: Sparse gamma rhythms arising through
clustering in adapting neuronal networks. PLoS Comput Biol 2011, 7(11):
e1002281.
4. Contreras D, Palmer L: Response to contrast of electrophysiologically defined
cell classes in primary visual cortex. J Neurosci 2003, 23(17):6936-6945.
5. Fries P: A mechanism for cognitive dynamics: neuronal communication
through neuronal coherence. Trends Cogn Sci (Regul Ed.) 2005, 9(10):474-480.
6. Lowet E, Roberts M, Hadjipapas A, Peter A, van der Eerden J, De Weerd P:
Input-Dependent Frequency Modulation of Cortical Gamma Oscillations
Shapes Spatial Synchronization and Enables Phase Coding. PLoS Comput
Biol 2015, 11(2):e1004072.
O11
Downstream changes in firing regularity following damage to the early
auditory system
Dan FM Goodman1*, Alain de Cheveign2, Ian M Winter3, Christian Lorenzi2
1
Department of Electrical and Electronic Engineering, Imperial College
London, London, UK; 2Laboratoire des Systmes Perceptifs, Centre National
de la Recherche Scientifique and cole Normale Suprieure, Paris, France;
3 ANFs could therefore lead to a disruption of the processing of temporal
Centre for the Neural Basis of Hearing, The Physiological Laboratory,
Department of Physiology, Development and Neuroscience, University of
Cambridge, Cambridge, UK
E-mail: d.goodman@imperial.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):O11
We demonstrate how an abstract mathematical model that approximates
a wide range of more detailed models can be used to make predictions
about hearing loss-related changes in neural behaviour.
One consequence of neurosensory hearing loss (noise-induced and aging-
related) is a reduced ability to understand speech, particularly in noisy
environments, and sometimes beyond what would be predicted from
reduced audibility. Indeed, this type of speech deficit can occur in listeners
with near-normal hearing thresholds [1]. A promising avenue of
investigation to explain this comes from experimental results in mice
showing that there can be a permanent loss of auditory nerve fibres (ANFs)
following temporary noise-induced hearing loss (i.e. when thresholds
return to normal after a few weeks) [2]. The downstream consequences of
this loss of fibres has not yet been systematically investigated (although see
[3]). We predict, using a theoretical analysis that applies to a wide range of
neural models, that the regularity of the spike trains of many neurons in the
cochlear nucleus (the next structure after the auditory nerve) will decrease
following a reduction in the number of input cells.
We present a mathematical analysis of the stationary behaviour of chopper
cells in the ventral cochlear nucleus, approximating them by a stochastic
process that is entirely characterised by its mean, standard deviation and
time constants. Furthermore, these constants can be straightforwardly
related to physiologically significant parameters including the number of
inputs and their average firing rates. From this approximation, we can
Figure 1(abstract O12) Estimated parameters of the Implicit Memory Model. Estimated values of h (weighting of implicit memory) as a function of
estimated values of s (percentage of internal noise) of Controls (blue) and Dyslexics (red). The optimal weighting h* is plotted in green. Gray area
depicts the confidence interval of 2.5% below the best performance. Inset. Median deviation from optimal weighting of previous trials. Dyslexics
deviation is larger than Controls (Mann-Whitney test, Z = 2.5, P < 0.01). Error bars denote inter-quartile range
Page 10 of 200
compute the regularity of the chopper cell spike trains measured as the
coefficient of variation of their interspike intervals (CV).
One simple prediction of this model is that when the intensity of a
stimulus changes, leading to a change in the average firing rate of the
ANF inputs, there will be a corresponding change in the regularity of the
chopper cell spike train. This prediction poses problems for the widely
used scheme for classifying chopper cells as sustained or transient based
on their ongoing CVs as it implies that the classification could be level-
dependent. We present a re-analysis of an existing experimental data set
that demonstrates that ongoing CV is indeed level-dependent in the
majority of chopper cells, and that in some cells (>7%) this leads to a
level-dependence in their classification.
Assuming a homeostatic regulation of long term firing rates, a loss of
ANFs will lead to an increase in the standard deviation of the stochastic
process and a consequent increase in the CV of the chopper cell. Some
choppers that were previously classified as sustained will become
transient, a substantial change in their behaviour that is highly likely to
disrupt auditory processing. While the function of chopper cells is still
debated, one suggested role is in the coding of temporal envelope [4],
which is widely agreed to be essential for understanding speech. Loss of
envelope, and consequently degrade speech intelligibility. We briefly
conclude by discussing the challenges of testing this hypothesis
experimentally.
References
1. Leger AC, Moore BCJ, Lorenzi C: Temporal and spectral masking release
in low-and mid-frequency regions for normal-hearing and hearing-
impaired listeners. J Acoust Soc Am 2012, 131(2):1502-1514.
2. Kujawa SG, Liberman MC: Adding insult to injury: cochlear nerve
degeneration after temporary noise-induced hearing loss. J Neurosci
2009, 29(45):14077-14085.
3. Lopez-Poveda EA, Barrios P: Perception of stochastically undersampled
sound waveforms: a model of auditory deafferentation. Front Neurosci
2013, 7:124.
4. Lorenzi C, Micheyl C, Berthommier F: Neuronal correlates of perceptual
amplitude-modulation detection. Hearing Research 1995, 90(1-2):, 219-227.
O12
Towards a computational model of Dyslexia
Sagi Jaffe-Dax*, Ofri Raviv, Nori Jacoby, Yonatan Loewenstein, Merav Ahissar
Edmond and Lily Safra Center for Brain Sciences, The Hebrew University of
Jerusalem, Jerusalem 91904, Israel
E-mail: sagi.jaffe@mail.huji.ac.il
BMC Neuroscience 2015, 16(Suppl 1):O12
Dyslexics are diagnosed for their poor reading skills. Yet, they
characteristically also suffer from poor verbal memory, and often from
poor auditory skills. We now hypothesize that dyslexia can be understood
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 2(abstract O12) Grand Average ERPs to the two-tone stimulation. A. Controls B. Dyslexics. Trials are sorted according to the trial type, Bias+
(where the impact of previous trials improves performance) and Bias- (where the impact of previous trials impairs performance). Controls P2 after the first
tone differs between the two trial types. Dyslexics evoked responses did not differ between the two trial types. Filled areas denote cross-subject SEM.
Small black rectangles under the plots denote the temporal location of the two tones in the trial. Middle insets. P2 region enlarged; Top right insets.
Single subject P2 area in Bias- versus Bias+ trials. The difference between the trial types is significantly larger among Controls than among Dyslexics
(Condition Group interaction: Mann-Whitney test, z = 2.5, P < 0.05)
computationally as a deficit in integrating prior information with noisy
observations. To test this hypothesis we analyzed performance in two
tones pitch discrimination task using a two-parameter computational
model. One parameter captures the internal noise in representing the
current event and the other captures the impact of recently acquired prior
information [1]. We found that dyslexics perceptual deficit can be
accounted for by inadequate adjustment of these components: low
weighting of their implicit memory in relation to their internal noise
(Figure 1). Using ERP measurements we found evidence for dyslexics
deficient automatic integration of experiments statistics (Figure 2). Taken
together, these results suggest that dyslexia can be understood as a well-
defined computational deficit.
Acknowledgements: This work was supported by Israel Science
Foundation (grant no. 616/11)
Reference
1. Raviv O, Ahissar M, Loewenstein Y: How recent history affects perception:
the normative approach and its heuristic approximation. PLoS Comput
Biol 2012, 8:e1002731.
O13
Investigating the effect of electrical brain stimulation using a
connectome-based brain network model
Tim Kunze1,2*, Alexander Hunold1, Jens Haueisen1, Viktor Jirsa3,4,5,
Andreas Spiegler3,4
1
Institute of Biomedical Engineering and Informatics, Ilmenau University of
Technology, Ilmenau, 98693, Germany; 2Max Planck Institute for Human
Cognitive and Brain Sciences, Leipzig, Germany; 3Aix-Marseille Universit,
Institut de Neurosciences des Systmes, Marseille, France; 4Institut National
de la Sant et de la Recherche Mdical, UMR_S 1106, 27 Bd Jean Moulin,
13385, Marseille Cedex 5, France; 5Centre National de la Recherche
Scientifique, 3, rue Michel-Ange, 75794, Paris, France
E-mail: tim.kunze@tu-ilmenau.de
BMC Neuroscience 2015, 16(Suppl 1):O13
Transcranial direct current stimulation (tDCS) leads to positive effects in
neurological and psychiatric diseases, such as depression, pain, or stroke,
which outlast the treatment itself. Although numerous influencing
stimulation parameters and factors are known, the mechanisms behind
tDCS remain unclear. To reveal the mechanisms tDCS started to be
considered to affect networks while (de)polarizing parts of the brain. We
study here the ability of tDCS as a tool to bias functional networks by
affecting the connections given the brain structure.
We used structural data, that is, a human connectome to construct a large-
scale brain network model of 74 cerebral areas, each described by a
Jansen and Rit model. The model was designed on the basis of the
neuroinformatics platform The Virtual Brain to account for reproducibility
of the simulations. The tDCS-induced currents on the cerebral areas were
Page 11 of 200
calculated using a finite element method model. Based on the dynamical
repertoire of an isolated area [1], we analyzed the brain activity, that is, the
spatiotemporal dynamics in terms of rhythms and baseline potentials
during rest, during tDCS, and the change between both.
We identified the network states during rest and catalogued all states for
further modeling studies. During tDCS, increased functional connectivity
was found among a set of scalp EEG sensors, as reported in measurements
[2], as well as among cerebral cortical areas (see Figure 1). Furthermore,
tDCS led to sharpened frequency spectra and increased (anode) or
decreased (cathode) power in the respective areas.
This study supports the notion that noninvasive brain stimulation is able to
bias brain dynamics by affecting the competitive interplay of functional
subnetworks. Our work constitutes a basis for further modeling studies to
test target-oriented manipulation of functional networks (e.g. through
adapted electrode montages) to improve pertinent treatment conditions.
Furthermore, our approach emphasizes the role of structural data such as
the network topology in emerging dynamics. Dynamics cannot necessarily
be predicted from the structure but we found the structure especially
important at transitions of network states.
References
1. Spiegler A, Kiebel SJ, Atay FM, Knoesche TR: Bifurcation analysis of neural
mass models: Impact of extrinsic inputs and dendritic time constants.
Neuroimage 2010, 52(3):1041-1058.
2. Polania R, Nitsche MA, Paulus W: Modulating functional connectivity
patterns and topological functional organization of the human brain
with transcranial direct current stimulation. Hum Brain Mapp 2011,
32(8):1236-1249.
O14
Closing the loop: optimal stimulation of C. elegans neuronal network
via adaptive control to exhibit full body movements
Julia Santos, Eli Shlizerman*
Department of Applied Mathematics, University of Washington, Seattle,
WA 98195, USA
E-mail: shlizee@uw.edu
BMC Neuroscience 2015, 16(Suppl 1):O14
The Caenorhabditis elegans (C. elegans) worm is a well-studied biological
organism model. The nervous system of C. elegans is particularly appealing
to study, since it is a tractable fully functional neuronal network for which
electro-physical connectivity map (connectome) is fully resolved [1,2]. In a
recent work, we succeeded in establishing a computational dynamical
model for the C. elegans nervous system and showed that robust
oscillatory movements in motor neurons along the body can be invoked
by constant current excitation of command sensory neurons (e.g. PLM
neurons associated with forward crawling) and that their activation
corresponds to low-dimensional Hopf bifurcation [3]. While these first
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract O13) New functional connections are established
during tDCS: among cortical areas, Panel A; and among scalp EEG
electrodes, Panel B
results validated the model, it is exciting to learn how the nervous system
transforms its oscillatory dynamics to the muscles to support robust full
body movements (e.g. forward crawling) [4]. Moreover, using methods
generically applicable to other neuronal circuits, it is intriguing to
understand the optimal sensory stimulations that cause these movements
to persist.
We explore these questions by modeling the C. elegans musculature as a
viscoelastic rod with discrete rigid segments [5], and map the neuronal
dynamics such that they activate the muscles and deform the rod (Fig.
1A). When motor neuron activity stimulates muscles [2], this activation is
translated into force applied to the rod, which moves in accordance with
Page 12 of 200
Figure 1(abstract O14) A: Structure of viscoelastic rod B: Viscoelastic
rod-based simulation of C. elegans crawling during PLM excitation,
videos available here C: Loop feeding transformed motor activity into
sensory neurons
the physical properties of C. elegans. By stimulating the command PLM
neurons, we establish for the first time that motor neuron dynamics are
indeed producing coherent oscillatory full body movements that
resemble forward crawling (Fig. 1B, videos available here: http://faculty.
washington.edu/shlizee/celegans/).
We utilize our computational full body model to determine the appropriate
sensory input for behavior, such as crawling, to persist after explicit external
stimulation (touch) has ceased, as observed in experiments [5]. Since such
persistence could be explained by a feedback loop between the
environment and sensory neurons (Fig. 1C), we propose an adaptive control
algorithm that extends existing recursive least squares-based algorithms
(e.g. FORCE [6]). Our algorithm finds weights for synaptic input using a low-
dimensional projection of motor neuron dynamics, and is capable of finding
sensory input patterns that will lead to the desired movement.
References
1. Varshney LR, Chen BL, Paniagua E, Hall DH, Chklovskii DB: Structural
Properties of the Caenorhabditis elegans Neuronal Network. PLoS Comput
Biol 2011, 7(2):1001066.
2. Sengupta P, Samuel ADT: Caenorhabditis elegans: a model system for
systems neuroscience. Curr Opin Neurobiol 2009, 19(6):1-7.
3. Kunert J, Shlizerman E, Kutz JN: Low-dimensional functionality of complex
network dynamics: Neurosensory integration in the Caenorhabditis
elegans connectome. Phys Rev E 2014, 89(5):052805.
4. McMillen T, Williams T, Holmes P: Nonlinear Muscles, Passive
Viscoelasticity and Body Taper Conspire To Create Neuromechanical
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Phase Lags in Anguilliform Swimmers. PLoS Comput Biol 2008,
4(8):1000157.
5. Backholm M, Ryu WS, Dalnoki-Veress K: Viscoelastic properties of the
nematode Caenorhabditis elegans, a self-similar, shear-thinning worm.
PNAS 2013, 110(12):4528-4533.
6. Sussillo D, Abbott LF: Generating Coherent Patterns of Activity from
Chaotic Neural Networks. Neuron 2009, 63(4):544-557.
O15
Collective information storage in multiple synapses enables fast
learning and slow forgetting
Michael J Fauth1*, Florentin Wrgtter1, Christian Tetzlaff2,3
1
Bernstein Center for Computaional Neuroscience, Third Physics Institute,
Gttingen, Germany; 2Max Planck Institute for Dynamics and Self-
Organization, Gttingen, Germany; 3Weizman Institute, Rehovot, Israel
E-mail: mfauth@gwdg.de
BMC Neuroscience 2015, 16(Suppl 1):O15
Most of the excitatory cortical synapses reside on dendritic spines.
Although these spines undergo a remarkably high turnover [1,2], they
have been shown to be involved in learning and long-term memory. Along
this line, it is unclear how information is preserved while its substrate
(synapses or spines) is permanently changing.
Here, we use a simple stochastic model of structural plasticity to
investigate this phenomenon : We assume a certain number of potential
synaptic locations from one neuron to another.
At those locations, synapses (spines) are created with a constant
probability and removed with a probability depending on the number of
existing synapses and the stimulation of the neurons. From these two
probabilities, the stationary distribution of the number of synapses
between two neurons can be calculated.
Experimental measurements of these stationary probability distributions in
the cortex show that the majority of connections has either zero or
multiple synapses while one or two contacts are very improbable [e.g.,
[3-5]]. Using information theoretic measures we show that, in our model,
such bimodal distributions enable information storage over time scales
many orders of magnitudes higher than the involved probabilities. Thus, in
this system the conflict of rapid spine turnover (probabilities) and long-
term memory is resolved by storing the information collaboratively in
multiple synapses.
In the following, we will consider the bimodal stationary distributions as
the working point of the system. Then, we can model external signals, as,
e.g., increased or decreased activities during learning, as changes of the
removal probabilities and stationary distributions (e.g., mediated by
synaptic plasticity [6]).
For instance, for learning signals resulting to unimodal stationary
distributions (only connected or only unconnected), we find that learning is
orders of magnitude faster than forgetting. Along this line, we observe that
retraining a task does not induce an increased overturn rate as during initial
training, which has been similarly observed for dendritic spines in vivo [1,2].
Our results clearly relate the difference in time scales to the shape of the
stationary distribution and therefore reveal the functional advantage of the
bimodal distribution found in experiment.
References
1. Yang G, Pan F, Gan WB: Stably maintained dendritic spines are associated
with lifelong memories. Nature 2009, 462:920-924.
2. Xu T, Yu X, Perlik AJ, Tobin WF, Zweig JA, Tennant K, et al: Rapid formation
and selective stabilization of synapses for enduring motor memories.
Nature 2009, 462:915-919.
3. Feldmeyer D, Egger V, Lbke J, Sakmann B: Reliable synaptic
connections between pairs of excitatory layer 4 neurones within a
single barrel of developing rat somatosensory cortex. J Physiol 1999,
521(1):169-190.
4. Feldmeyer D, Lbke J, Silver RA, Sakmann B: Synaptic connections
between layer 4 spiny neurone-layer 2/3 pyramidal cell pairs in juvenile
rat barrel cortex: physiology and anatomy of interlaminar signalling
within a cortical column. J Physiol 2002, 538(Pt 3):803-822.
5. Hardingham NR, Hardingham GE, Fox KD, Jack JJB: Presynaptic efficacy
directs normalization of synaptic strength in layer 2/3 rat neocortex
after paired activity. J Neurophysiol 2007, 97(4):2965-2975.
Page 13 of 200
6. Fauth M, Wrgtter F, Tetzlaff C: The formation of multi-synaptic
connections by the interaction of synaptic and structural plasticity and
their functional consequences. PLOS Comput Biol 2015, 11(1):e1004031.
O16
Limited range correlations, when modulated by firing rate, can
substantially improve neural population coding
Joel Zylberberg1*, Jon Cafaro2, Maxwell Turner3, Fred Rieke3,4,
Eric Shea-Brown1
1
Department of Applied Mathematics, University of Washington, Seattle, WA
98195, USA; 2Department of Biology, Duke University, Durham, NC 27708,
USA; 3Department of Physiology and Biophysics, University of Washington,
Seattle, WA 98195, USA; 4Howard Hughes Medical Institute, University of
Washington, Seattle, WA 98195, USA
E-mail: joelzy@uw.edu
BMC Neuroscience 2015, 16(Suppl 1):O16
Neural activities are unreliable indicators of features of the external world
[1], and an open question is how our nervous systems function robustly in
the presence of this noise. One possibility arises from the observation that
variability is often correlated between neurons [2], leading to the
important theoretical question of when and how noise correlations affect
neural population codes. Much work has investigated this issue, leading to
impressive insights about how the relationship between the statistical
structures of signals vs noise affects neural population coding. Despite this
progress, an important issue has been largely overlooked by the field: that
of firing-rate-dependent correlations. Notably, the same pair of neurons
can display different noise correlations in response to different stimuli;
those correlations coefficients typically increase with increases in the
neurons firing rates [2,3].
In this paper, we investigate the role of so-called limited range
correlations on population codes, either in the presence, or the absence of
rate-modulation of the correlation coefficients. Limited-range correlations
are frequently-observed population-wide correlation structures in which
cells with similar tuning curves have positive noise correlations, and the
correlations decrease with decreasing tuning curve similarity [2,4]. These
patterns of noise correlation are typically harmful to population coding [5]
(yielding worse population coding performance than would be obtained
with the same tuning curves and noise variances for all cells, but no noise
correlations); these effects are somewhat dependent on the degree of
heterogeneity in the populations tuning curves [6,7].
Experimentally reported noise correlations are usually averaged over stimuli,
thereby masking any stimulus dependence. Herein, we will demonstrate that,
when correlation coefficients increase with the product of mean neural firing
rates (as in [2,3]), the stimulus-averaged correlation coefficients will display
limited-range structure. When the rate dependence of these correlations is
ignored, those correlations appear to be quite harmful to the population
code, in accordance with previous theoretical work [5]. Surprisingly, when the
rate dependence is taken into account, the correlations can yield much
better population coding than would be obtained in the presence of
uncorrelated noise. These effects persist for either homogeneous or
heterogeneous sets of neural tuning curves. One prior study [8] also found
that rate-dependent correlations can have very different impacts on
population codes than can rate-independent ones, but did not make the
connection between firing-rate-dependent correlations and the frequently
observed limited-range correlation structure. Overall, our results emphasize
that, for understanding the impact of limited-range correlations on neural
population coding, the firing-rate dependence of those correlations is a
potentially important consideration. Thus, it is important to report not only
stimulus-averaged correlation coefficients, but also the relationship between
those correlation coefficients and the neural firing rates.
References
1. Faisal AA, Selen LPJ, Wolpert DM: Noise in the nervous system. Nat Rev
Neurosci 2008, 9(4):292-303.
2. Cohen MR, Kohn AK: Measuring and interpreting neuronal correlations.
Nat Neurosci 2011, 14(7):811-819.
3. de la Rocha J, Doiron B, Shea-Brown E, Josic K, Reyes A: Correlation between
neural spike trains increases with firing rate. Nature 2007, 448:802-806.
4. Zohary E, Shadlen MN, Newsome WT: Correlated neuronal discharge rate
and its implications for psychophysical performance. Nature 1994,
370(6485):140-143.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
5. Averbeck BB, Latham PE, Pouget A: Neural correlations, population coding
and computation. Nat Rev Neurosci 2006, 7(5):358-366.
6. Ecker AS, Berens P, Tolias AS, Bethge M: The effect of noise correlations in
populations of diversely tuned neurons. J Neurosci 2011,
31(40):14272-14283.
7. Shamir M, Sompolinsky H: Implications of neuronal diversity on
population coding. Neural Comput 2006, 18(8):1951-1986.
8. Josic K, Shea-Brown E, Doiron B, de la Rocha J: Stimulus-dependent
correlations and population codes. Neural Comput 2009, 21(10):2774-2804.
O17
Multiple mechanisms of theta rhythm generation in a model of the
hippocampus
Ali Hummos1,2*, Satish S Nair3
1
Department of Health Informatics, University of Missouri, Columbia, MO
65211, USA; 2Department of Psychiatry, University of Missouri, Columbia, MO
65211, USA; 3Department of Electrical & Computer Engineering, University of
Missouri, Columbia, MO 65211, USA
E-mail: hummosa@gmail.com
BMC Neuroscience 2015, 16(Suppl 1):O17
Hippocampal theta oscillations (4-12 Hz) are consistently recorded during
memory tasks and spatial navigation. While computational models
suggested specific mechanisms for theta generation, experimental
inactivation of these mechanisms did not disrupt theta, precluding
definitive conclusions about their roles. We investigated this discrepancy
using a biophysical model of the hippocampus that included several of
the components implicated in rhythm generation, all constrained by prior
experimental results. The CA3 network model included recurrently
connected pyramidal cells, and inhibitory basket cells (BC) and oriens-
lacunosum moleculare (OLM) cells. The model was developed by
matching experimental results characterizing neuronal firing patterns,
synaptic dynamics, short-term synaptic plasticity and the three-
dimensional organization of the hippocampus. The model revealed four
mechanisms that generated theta oscillations: intrinsic theta resonance of
pyramidal cells, recurrent connections between them, coupling between
OLM and pyramidal cells, and, as a novel finding, the correlated input
from entorhinal cortex. Consistent with experimental results, inactivation
of any single mechanism did not disrupt the rhythm. Another novel
finding was that the low and high cholinergic states differentially
recruited theta generating mechanisms. Atropine -sensitive and -resistant
forms of theta, however, corresponded to theta generated during low
and high levels of network excitation, respectively. These findings
provided an alternative interpretation of the atropine-based classification
of theta oscillations, and suggested that the theta rhythm is an intrinsic
property of the network. Any experimental manipulation or brain state
that enhances or suppresses excitation might also, therefore, non-
specifically enhance or suppress theta oscillations.
O18
Modelling phase precession in the hippocampus
Angus Chadwick1*, Mark van Rossum1, Matthew Nolan2
1
Institute for Adaptive and Neural Computation, University of Edinburgh,
Edinburgh, UK; 2Centre for Integrative Physiology, University of Edinburgh,
Edinburgh, UK
E-mail: angus.chadwick87@gmail.com
BMC Neuroscience 2015, 16(Suppl 1):O18
The activity of cells in the rodent hippocampus is strongly modulated by
both the location of the animal and the ongoing theta oscillation. Place
cells, but not interneurons, show a strong spatial modulation of their firing
rates, while both place cells and interneurons exhibit phase precession, a
phenomenon whereby they spike at a faster frequency than the LFP theta
oscillation, causing their spikes to shift to an earlier phase of this rhythm
on each successive cycle [1-3]. Despite extensive research into this
phenomenon, the mechanisms underlying phase precession remain
unclear.
Place cells and interneurons are reciprocally connected in the CA1 region
of the hippocampus. The interneurons receive pacemaker input from the
Page 14 of 200
medial septum, which entrains theta oscillations in the circuit. We tested
whether a minimal model based on this architecture could produce
phase precession in place cells and interneurons. Specifically, we
simulated a single place cell and interneuron, which interact synaptically.
The interneuron was driven with a constant depolarising input which
generates tonic spiking, as well as a weak theta oscillation which entrains
this spiking activity. The place cell received a depolarising input which is
only active at a certain location in the environment, representing the
place field.
We found that phase precession in both the place cell and interneuron
emerges naturally in this model. When the animal is outside of the place
field, the interneuron is fully entrained to the pacemaker theta oscillation,
and the place cell is rhythmically inhibited, resulting in subthreshold
theta oscillations. When the animal enters the place field, the place cell
begins to spike, which perturbs the interneuron and causes it to
transiently fire at a frequency higher than the pacemaker input. In turn,
the spiking of the interneuron entrains the place cell, generating phase
precession in the coupled pair.
Generalisation of this model to the network level reveals important
constraints. In particular, as there are far fewer interneurons than place
cells in CA1, it is necessary that the same interneuron is coupled to
multiple place cells. In our model, a single interneuron can successfully
couple to multiple place cells to generate phase precession, provided
that their place fields do not overlap. This poses constraints on the
possible place field mappings in such a network, and places limits on the
fraction of place cells which can be active in a single environment. When
working within these limits, the network can flexibly generate phase
precession, both in linear and open environments, across a vast number
of distinct place field mappings.
Our model has several advantages over existing models. First, our model
generates phase precession through the intrinsic dynamics of the circuit,
without the need for velocity controlled oscillators upstream. Second, our
model can generate omnidirectional phase precession in open
environments, without additional inputs from head direction cells. Finally,
our model generates phase precession independently in each cell, and
therefore allows spatial representations to be flexibly remapped without
detriment to the temporal coding of spatial trajectories in the population
[4].
References
1. OKeefe J, Recce M: Phase relationship between hippocampal place units
and the EEG theta rhythm. Hippocampus 1993, 3(3):317-330.
2. Maurer AP, Cowen SL, Burke SN, Barnes CA, McNaughton BL: Phase
precession in hippocampal interneurons showing strong functional
coupling to individual pyramidal cells. J Neurosci 2006,
26(52):13485-13492.
3. Geisler C, Robbe D, Zugaro M, Sirota A, Buzsaki G: Hippocampal place cell
assemblies are speed-controlled oscillators. PNAS 2007,
104(19):8149-8154.
4. Chadwick A, van Rossum MCW, Nolan MF: Independent theta phase
coding accounts for CA1 population sequences and enables flexible
remapping. Elife 2015, 2:4.
O19
Self-organization to sub-criticality
V Priesemann1,2
1
Department of Nonlinear Dynamics, Max Planck Institute for Dynamics and
Self-Organization, Gttingen, Germany; 2Bernstein Center for Computational
Neuroscience, Gttingen, Germany
E-mail: viola@nld.ds.mpg.de
BMC Neuroscience 2015, 16(Suppl 1):O19
Human brains possess sophisticated information processing capabilities,
which rely on the interactions of billions of neurons. However, it is
unclear how these capabilities arise from the collective spiking dynamics.
A popular hypothesis is that neural networks assume a critical state [1,2],
because in models criticality maximizes information processing
capabilities [3,4]. However, it has been largely overlooked that criticality
in neural networks also comes with the risk of spontaneous runaway
activity [5], which has been linked to epilepsy. Does the brain indeed
assume a critical state, despite the risk of instability? To revisit this
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 15 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

question, we analyzed spiking activity from awake animals, instead of

more coarse measures of neural activity (population spikes, LPF, EEG,
BOLD) as in most previous studies. In all recordings (rats hippocampus,
cats visual cortex, and monkey prefrontal cortex), spiking activity
resembled a sub-critical state, not criticality proper [6]. We confirmed
these results using a novel mathematical approach that is robust to
subsampling effects [7] [see Wilting & Priesemann, conference
proceedings CNS 2015]. While self-organization to criticality has been
widely studied (e.g.[5,8]), it is unclear what mechanism allows self-
organize to sub-criticality instead. Here, we demonstrate that homeostatic
plasticity [9] assures that networks assume a slightly sub-critical state,
independently of the initial configuration. Surprisingly, increasing the
external input (stimuli) altered the set-point of the network to a more
sub-critical state. Our results suggest that homeostasis allows the brain to
maintain a safety margin to criticality. Thereby the brain may lose
processing capability, but avoids instability.
References
1. Beggs JM, Plenz D: Neuronal avalanches in neocortical circuits. J Neurosci
2003, 23:11167-11177.
2. Priesemann V, Valderrama M, Wibral M, Le Van Quyen M: Neuronal
Avalanches Differ from Wakefulness to Deep Sleep-Evidence from
Intracranial Depth Recordings in Humans. PLoS Comput Biol 2013, 9:
e1002985.
3. Boedecker J, Obst O, Lizier JT, Mayer NM, Asada M: Information processing
in echo state networks at the edge of chaos. Theory Biosci 2012,
131:205-213.
4. Bertschinger N, Natschlger T: Real-time computation at the edge of
chaos in recurrent neural networks. Neural Comput 2004, 16:1413-1436.
5. Bak P, Tang C, Wiesenfeld K: Self-organized criticality: An explanation of
1/f noise. Phys Rev Lett 1987, 59:381-384.
6. Priesemann V, Wibral M, Valderrama M, Prpper R, Le Van Quyen M,
Geisel T, et al: Spike avalanches in vivo suggest a driven, slightly Figure 1(abstract O20) A. Spatial attractors. For each value of the
subcritical brain state. Front Syst Neurosci 2014, 8:108. global (G) or local (G_local) coupling parameter, correlation with
7. Priesemann V, Munk MH, Wibral M: Subsampling effects in neuronal in-strength (blue points) and with s-core (red points) for ten
avalanche distributions recorded in vivo. BMC Neurosci 2009, 10:40. different initial conditions. The blue square indicates the critical
8. Levina A, Herrmann JM, Geisel T: Dynamical synapses causing self- interval. B. Values for the beginning and end of the critical range as a
organized criticality in neural networks. Nat Phys 2007, 3:857-860. function of G or G_local
9. Turrigiano G: Homeostatic synaptic plasticity: local and global
mechanisms for stabilizing neuronal function. Cold Spring Harb Perspect
Biol 2012, 4:a005736.
stimulation, using an excitable neural mass model. We investigated the
effect of heterogeneous and homogeneous connectivity on large-scale
brain dynamics by different numbers of regions in the parcellation (70 to
O20 2240) and by varying the local connectivity coupling strength. To
Large-scale brain dynamics: effect of connectivity resolution introduce experimental data (i.e., structural and diffusion MRI) into TVB we
Timothe Proix*, Andreas Spiegler, Viktor K Jirsa tackled issues such as surface downsampling (for achieving moderate
Aix Marseille Universit, Inserm, INS UMR_S 1106, 13005, Marseille, France simulation times) and mapping between surface and parcellation (to
E-mail: timothee.proix@etu.univ-amu.fr consistently use heterogeneous and homogeneous connectivity) by
BMC Neuroscience 2015, 16(Suppl 1):O20 developing the Surface and Connectivity Reconstruction with an Imaging
Pipeline for TVB Simulations, short SCRIPTS [4].
Large-scale brain dynamics recently started to be modeled numerically When considering slow dynamics, the major fiber bundles best reflected
based on both heterogeneous large-scale networks build from diffusion MRI, in the coarsest parcellation appeared to be mainly responsible for the
that is, a connectome, and local homogeneous connectivity kernel emergence of the network attractors with limited changes over different
representing intracortical synaptic connections. However, topological parcellations and different local coupling strengths. For fast dynamics,
properties of a connectome can significantly change with resolution and new qualitative solutions appeared, but only in the presence of delays.
parcellation [1]. Furthermore, the sampling of the cerebral surfaces, resulting Acknowledgements: This research has been supported by the Brain
in a geometric model, and, in this way, the local connectivity kernel play Network Recovery Group through James S. McDonnell Foundation and
crucial roles in the formation of spatial patterns on the cerebral surfaces as funding from the European Union Seventh Framework Programme (FP7-
well as on the sensor level (e.g., EEG electrodes) by a forward calculation [2]. ICT BrainScales and Human Brain Project (grant no. 60402))
However, the effect of sampling and parcellation on modeling brain References
dynamics has not been studied so far. Here, we investigate qualitatively and 1. Zalesky A, Fornito A, Harding IH, Cocchi L, Ycel M, Pantellis C, Bullmore ET:
quantitatively: (i) how different parcellation resolutions affect the dynamics Whole-brain anatomical networks: does the choice of node matter?
of the network; and (ii) how the local connectivity affects the network Neuroimage 2010, 50(3):970-983.
dynamics. To do so, we used the neuroinformatics platform for large-scale 2. Spiegler A, Jirsa V: Systematic approximation of neural fields through
brain simulations, called The Virtual Brain (TVB) [3] and developed a networks of neural masses in the virtual brain. Neuroimage 2013, 83:704-725.
preprocessing pipeline to incorporate experimental data (e.g., structural MRI, 3. Sanz-Leon P, Knock SA, Woodman MM, Domide L, Mersmann J,
diffusion-weighted MRI) in TVB [4]. McIntosh AR, Jirsa VK: The Virtual Brain: a simulator of primate brain
We prepared ten individual models based on ten randomly selected network dynamics. Frontiers in Neuroinformatics 2013, 7:1-23.
subjects from the Human Connectome Project dataset [5]. For each 4. SCRIPTS. [https://github.com/timpx/scripts].
individual model we performed simulations under two conditions during 5. Van Essen DC, Smith SM, Barch DM, Behrens TEJ, Yacoub E, Ugurbil K: The
rest: (i) noise driven, using a bistable neural mass model, and (ii) after WU-Minn Human Connectome. Neuroimage 2013, 80:62-79.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
POSTER PRESENTATIONS
P1
Computational modeling of heterosynaptic plasticity in the
hippocampus
Peter Jedlicka1*, Lubica Benuskova2,4, Wickliffe C Abraham3,4
1
Institute of Clinical Neuroanatomy, Neuroscience Center, Goethe University
Frankfurt, Frankfurt am Main, Germany; 2Department of Computer Science,
University of Otago, Dunedin, New Zealand; 3Department of Psychology
University of Otago, Dunedin, New Zealand; 4Brain Health Research Centre,
University of Otago, Dunedin, New Zealand
E-mail: jedlicka@em.uni-frankfurt.de
BMC Neuroscience 2015, 16(Suppl 1):P1
Hippocampal long-term potentiation (LTP) and long-term depression (LTD)
are central synaptic mechanisms of learning and memory. Here we use
compartmental models of hippocampal granule cells to better understand
LTP and heterosynaptic LTD which have been reported in the dentate gyrus
of awake rats [1]. Our simulations indicate that LTP and heterosynaptic LTD
can be explained by a spike-timing-dependent plasticity (STDP) rule
combined with a fast Bienenstock-Cooper-Munro (BCM)-like metaplasticity
rule [2-5]. We study the interaction between these plasticity rules and
ongoing pre- and postsynaptic activity. Our models are able to account for
the experimentally observed degree of LTP and heterosynaptic LTD induced
by various plasticity-inducing protocols.
Acknowledgements: The work was supported by a Young Investigators
Grant (from the faculty of medicine Goethe-University to P.J.) and by a
BMBF grant (Germany-USA Collaboration in Computational Neuroscience
to P.J., No. 01GQ1203A)
References
1. Bowden JB, Abraham WC, Harris KM: Differential effects of strain, circadian
cycle, and stimulation pattern on LTP and concurrent LTD in the
dentate gyrus of freely moving rats. Hippocampus 2012, 22(6):1363-1370.
2. Abraham WC: Metaplasticity: tuning synapses and networks for plasticity.
Nat Rev Neurosci 2008, 9(5):387.
3. Benuskova L, Abraham WC: STDP rule endowed with the BCM sliding
threshold accounts for hippocampal heterosynaptic plasticity. J Comp
Neurosci 2007, 22(2):129-133.
4. Abraham WC, Logan B, Wolff A, Benuskova L: Heterosynaptic LTD in the
dentate gyrus of anesthetized rat requires homosynaptic activity.
J Neurophysiol 2007, 98(2):1048-1051.
5. Benuskova L, Jedlicka P: Computational modeling of long-term
depression of synaptic weights: insights from STDP, metaplasticity and
spontaneous activity. Neural Network World 2012, 22(2):161-180.
P2
The sensori-motor model of the hippocampal place cells
Anu Aggarwal
Electrical and Computer Engineering Department, University of Maryland,
College Park, MD, 20742, USA
E-mail: aaagganu@gmail.com
BMC Neuroscience 2015, 16(Suppl 1):P2
The hippocampal formation contains the head direction cells, the grid cells
and the place cells which work as an internal GPS for the brain. The head
directions cells can sense the direction in which the animal is moving,
based on which the entorhinal grid cells fire at regular intervals as the
animal is at the corners of an equilateral triangle and the hippocampal
place cells [1] fire when the animal is at a place in the environment.
Several mathematical models [2] have been proposed to explain firing
pattern of the place cells, most of which consider place cell firing to be the
result of integration of either sensory or processed motor inputs (received
via the grid cells). Only the oscillatory interference model mentions the
role of both sensory and motor inputs in place cell firing but does not
explain how this information is remembered. However, empirical
observations [2-6] indicate a role for both the sensory and the grid cell
inputs in place cell firing and a one to one correspondence between the
grid and the place cell firing patterns. Anatomical evidence [7] indicates
that an area of the medial entorhinal cortex is primarily connected to a
similar area of the hippocampus along the dorso-ventral axis. All the above
Page 16 of 200
is not possible if the place cell firing depended on integration of inputs
from several grid cells or on integration of either sensory or motor inputs
alone as proposed by the current models. Moreover, the animals without
binocular vision, in whom most place cell recordings have been done,
ought to have different mechanism for mapping the environment than
those with binocular vision. Due to lack of depth perception, the former
could be using the matrix formed by grid cells to hang the objects as seen
by their eyes to get an accurate estimate of position. Therefore, we
propose that the place cells integrate both sensory and motor inputs to
them in a Bayes optimal manner (Fig. 1). Also each place cell is connected
to only one grid cell. As the animal enters an environment, the place cell
resets. While the animal moves around in the environment, the grid cell
firing causes movement along the ring of intermediate cells which wraps
around. As soon as the sensory inputs identify the place to be of
significance and cause a place cell to fire, by Hebbian learning, synapses
develop between the place cell and the intermediate neuron which is
firing concurrently in the ring. These synaptic connections help the animal
to retrieve the location correctly during future encounters. This model
offers improvements over the prior models as it explains the empirical
observations more closely.
References
1. OKeefe J, Conway DH: Hippocampal Place Units in the freely moving rat:
Why they fire where they fire. Exp Brain Res 1978, 31:573-590.
2. Hasselmo ME: How we remember MIT press, 1 2012.
3. OKeefe J, Burgess N: Dual phase and rate coding in hippocampal place
cells: theoretical significance and relationship to entorhinal grid cells.
Hippocampus 2005, 15:853-866.
4. Wills TJ, Cacucci F, Burgess N, OKeefe J: Development of the
Hippocampal Cognitive Map in Preweanling Rats. Science 2010,
328:1573-6.
5. Chen G, King JA, Burgess N, OKeefe J: How vision and movement
combine in the hippocampal place code. PNAS 2013, 110(1):378-383.
6. Hafting T, Fyhn M, Bonnevie T, Moser MB, Moser EI: Hippocampus
independent phase precession in entorhinal grid cells. Nature 2008,
453:1248-1252.
7. Witter MP, Groenewegen HJ, Lopes da Silva FH, Lohman AHM: Functional
organisation of the extrinsic and intrinsic circuitry of the
parahippocampal region. Prog Neurobiol 1989, 33:161-253.
P3
Short desynchronization epochs in neural synchronization: detection,
mechanisms, and functions
Leonid L Rubchinsky1,2*, Sungwoo Ahn1,3
1
Department of Mathematical Sciences, Indiana University Purdue University
Indianapolis, Indianapolis, IN, USA; 2Stark Neurosciences Research Institute,
Indiana University School of Medicine, Indianapolis, IN, USA; 3Present
address: School of Mathematical and Statistical Sciences, Arizona State
University, Tempe, AZ, USA
E-mail: leo@math.iupui.edu
BMC Neuroscience 2015, 16(Suppl 1):P3
Synchrony of neural oscillations is believed to play important role in a
variety of functions and dysfunctions of neural systems. However, neural
synchrony can hardly be perfect for prolong intervals of time. If there is a
significant synchrony strength observed over sufficiently long time interval,
the imperfect nature of synchrony implies that for some intervals of time
synchrony may be stronger, while for some intervals of time in may be
weaker. If neural synchrony is to play some functional role, this role may
depend on how synchrony is patterned in time. Roughly speaking, few long
intervals of synchronous (or nonsynchronous) dynamics may be functionally
different from many short intervals, although the synchrony strength may
be the same on the average.
Recent developments in time-series analysis allowed exploring the
patterning of synchrony in time on very short time-scales. If there is some
synchrony level is present on the average, than one can look at each cycle
of oscillations and detect of signals are synchronous or not resulting in a
quantitative description of succession of episodes of synchronization and
desynchronization [1]. We applied these methods to analyze neural signals
of various natures: synchronization of single units and LFPs in subthalamic
nucleus of Parkinsonian patients [2], synchronization of EEG signals in
healthy humans [3], synchronization of LFPs within and between prefrontal
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P2) Sensorimotor model of the hippocampal place cells. A) Hexagonal grid cell firing gets integrated on to B) the ring of
intermediate neurons. C) Together sensory and motor inputs are integrated in a Bayes optimal manner on the place cell
and hippocampal circuits in normal rats and rats experiencing repetitive
psychostimulant injections [4]). The temporal patterning of the neural
synchrony was sensitive to behavioral states, drug status etc. However, all
these studies yielded one important qualitative results: neural oscillations
were observed to go out of synchrony primarily for a short time interval. They
may go out of synch very frequently (leading to low synchrony levels), but all
these different neural systems exhibit mostly very short desynchronization
intervals (different from rare prolongs desynchronizations and a spectrum of
possibilities in between, none of which were experimentally observed). On
the contrary, non-neural systems can exhibit longer desynchronization
durations [1,5].
These experimental results suggest that short desynchronization dynamics
may be universal for synchronized activity of neural systems. Here we
review the data analysis techniques and results of their application to
experimental data. We then consider the potential mechanisms of the
apparent universality of short desynchronization dynamics and its potential
functional advantages. We consider simple neural circuits of Morris-Lecar-
type neurons and identify conditions imposed on the parameter values to
result in both short desynchronization dynamics and dynamics exhibiting
longer and rarer desynchronized intervals. By maintaining the same levels of
average synchrony (and the same frequency of oscillations) we show that
fast conductances promote short desynchronization dynamics observed in
experiments. By subjecting these model networks to oscillatory synaptic
input, we found that systems with short desynchronization dynamics require
weaker input signal in order to synchronize with this input. We finally
conjecture that properties of ionic conductances responsible for short
desynchronization dynamics may make neural systems more adaptable, as
they are more efficient in generating transient synchronous patterns
(synchronous assemblies) in response to synaptic or sensory inputs.
References
1. Ahn S, Park C, Rubchinsky LL: Detecting the temporal structure of
intermittent phase locking. Phys Rev E 2011, 84(1-2):016201.
2. Park C, Worth RM, Rubchinsky LL: Fine temporal structure of beta
oscillations synchronization in subthalamic nucleus in Parkinsons
disease. J Neurophysiol 2010, 103(5):2707-2716.
3. Ahn S, Rubchinsky LL: Short desynchronization episodes prevail in the
synchronous dynamics of human brain rhythms. Chaos 2013,
23(1):013138.
4. Ahn S, Rubchinsky LL, Lapish CC: Dynamical reorganization of synchronous
activity patterns in prefrontal cortex - hippocampus networks during
behavioral sensitization. Cerebral Cortex 2014, 24(10):2553-2561.
5. Rubchinsky LL, Ahn S, Park C: Dynamics of synchronization-
desynchronization transitions in intermittent synchronization. Frontiers in
Computational Physics 2014, 2:38.
Page 17 of 200
P4
Robust estimation of millisecond timescale synchrony under
nonstationary conditions and its physiological interpretation
Jonathan Platkiewicz1*, Kamran Diba2, Pascale Quilichini3,4, Gyrgy Buzsaki5,
Asohan Amarasingham1
1
Department of Mathematics, City College, City University of New York, New
York, NY 10031, USA; 2Department of Psychology, University of Wisconsin-
Milwaukee, Milwaukee, WI 53201, USA; 3Aix Marseille Universit, Institut des
Neurosciences des Systmes, Marseille, France; 4Inserm, UMR_S 1106, 27 Bd
Jean Moulin, 13385 Marseille Cedex 5, France; 5Neuroscience Institute,
New York University, New York, NY 10016, USA
E-mail: jonathan.platkiewicz@gmail.com
BMC Neuroscience 2015, 16(Suppl 1):P4
Motivated by the observation of millisecond-timescale synchrony among
hippocampal neurons in high-density extracellular recordings of freely
moving rats [1], we developed and applied a model for detecting
synchrony under such conditions. We calibrated the model using data of a
putative monosynaptic connection, obtained in paired extracellular-
intracellular recordings in anesthetized rats [2].
There has been a great deal of interest in detecting synchrony, and more
generally fine-temporal correlation, in the firing activities of pairs of neurons.
In certain settings, a common question is whether this fine-temporal
structure is supposed to reflect either a direct synaptic connection between
these cells ("the monosynaptic connection hypothesis), or a larger network
organization in which these cells are embedded ("the common modulation
hypothesis). Here, we studied a specific related problem: what is a proper
estimate of the amount of synchrony between two spike trains?
We modeled the problem based on a technique we call separation of time
scale model. Among synchronous spikes, a spike can belong either to a
fast timescale process, or to a slow timescale process, called background
activity; non-synchronous spikes belong to background activity. The
definition of timescale for these processes is based on conditional
modeling tools, as developed in previous studies [3], and is designed to
accommodate even extreme forms of nonstationarity in the underlying
spike processes. Under this framework, we can either test the hypothesis
that the activities of these two cells are uncorrelated on a fine-timescale, or
estimate the number of synchronous spikes due to fine-timescale
processes. It was argued however that, in the context of related models,
resulting statistical hypothesis tests have low power and estimators of
synchrony are biased [4]. An alternative method was proposed and
observed to offer a high power test and unbiased estimation in numerical
simulations. We studied this alternative approach and framed it within our
separation-of-timescale model. We develop hypothesis tests and unbiased
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
estimators and interpret and clarify previous results in this context. We
applied the developed method to large-scale recordings of neuronal
populations in the cornu ammonis 1 (CA1) and CA3 regions of the
hippocampus of freely moving rats [1]. Finally, we also simulated a simple
biophysical model of a monosynaptic connection, and demonstrated the
relevance of our method in this mechanistic context.
Nevertheless, this method is based solely on an analysis of spike trains,
which give a restricted representation of a cells `true activity and result
from numerous steps of statistical inference (e.g. spike sorting). Any
conclusion based on such analysis should be therefore trusted with caution.
We examined an experimental dataset where a fine-temporal structure was
identified using paired extracellular-intracellular in vivo recordings, in the
enthorinal cortex of the anaesthetized rat [2]. This dataset allowed us to
calibrate our spike train analysis method (i.e. supervised learning), and to
advance existing methodologies for identifying functional connectivity from
in vivo recording.
Acknowledgements: This work was supported by NIH grant R01MH102840.
References
1. Diba K, Amarasingham A, Mizuseki K, Buzsaki G: Millisecond Timescale
Synchrony among Hippocampal Neurons. J Neurosci 2014,
34(45):14984-14994.
2. Quilichini P, Sirota A, Buzsaki G: Intrinsic circuit organization and theta-
gamma oscillation dynamics in the entorhinal cortex of the rat.
J Neurosci 2010, 30(33):11128-11142.
3. Amarasingham A, Harrison MT, Hatsopoulos NG, Geman S: Conditional
modeling and the jitter method of spike resampling. J Neurophysiol 2012,
107(2):517-531.
4. Stark E, Abeles M: Unbiased estimation of precise temporal correlations
between spike trains. J Neurosci Meth 2009, 179(1):90-100.
P5
Effects of a reduced efficacy of the KCC2 co-transporter in temporal
lobe epilepsy: single neuron and network study
Anatoly Buchin1,2*, Gilles Huberfeld3,4, Richard Miles5, Anton Chizhov6,
Boris Gutkin1,7
1
cole normale suprieure, Laboratoire des Neurosciences Cognitives, Group
for Neural Theory, Paris, France; 2Peter the Great St.-Petersburg Polytechnic
University, St.-Petersburg, Russia; 3Neurophysiology Department, Pitie-
Salpetriere Hospital, UPMC, Paris, France; 4Epilepsie de lEnfant et Plasticit
Crbrale, INSERM U1129, Paris, France; 5Institut du Cerveau et de la Moelle
Epiniere, Cortex et Epilepsie Group, Paris, France; 6Ioffe Physical Technical
Institute, Computational Physics Laboratory, St.-Petersburg, Russia; 7Higher
School of Economics, Moscow, Russia
E-mail: anat.buchin@gmail.com
BMC Neuroscience 2015, 16(Suppl 1):P5
Epilepsy is one of the most common neurological disorders. Seizures in
about 40% of patients with temporal lobe epilepsies are pharmaco-
resistant [1]. In surgically removed hippocampal tissue from these patients,
the KCC2 cotransporter is absent or non-functional in about 20 % of
subicular pyramidal cells [2]. KCC2 normally assures the maintenance of
low intra-neuronal chloride levels [3] and also regulates potassium levels
[4]. Chloride concentration changes in the remaining pyramidal cells due
to intensive GABAergic input during seizures could reverse the effects of
GABA currents from inhibitory to excitatory [5,6]. Such changes may
shift a pyramidal cell into a periodic bursting regime associated with ictal
discharges. Using a detailed biophysical model of a single cell
incorporating these mechanisms of ionic homeostasis and a neural
network model, we show that decreasing the activity of KCC2 pump leads
to repetitive seizure-like firing in the pathologic network due to increased
extracellular potassium and intracellular chloride (Fig. 1). This model
provides insights into how a dysregulation of pyramidal cell chloride
homeostasis due to reduced levels of the KCC2 cotransporter may lead to
seizures in the epileptic human subiculum.
Acknowledgements: This work has been supported by the following
grants: ANR-10-LABX-0087 IEC, ANR-10-IDEX-0001-02 PSL, ERC-322721,
FRM FDT20140930942. Especially we would like to thank Giri Krishnan for
useful discussions.
References
1. Beghi E, Berg A, Carpio A, Forsgren L, Hesdorffer DC, Hauser WA, Tomson T:
Comment on epileptic seizures and epilepsy: definitions proposed by
Page 18 of 200
the International League Against Epilepsy (ILAE) and the International
Bureau for Epilepsy (IBE). Epilepsia 2005, 46(10):1698-1699.
2. Huberfeld G, Wittner L, Clemenceau S, Baulac M, Kaila K, Miles R, Rivera C:
Perturbed chloride homeostasis and GABAergic signaling in human
temporal lobe epilepsy. J Neuroscience 2007, 27(37):9866-9873.
3. Blaesse P, Airaksinen MS, Rivera C, Kaila K: Cation-chloride cotransporters
and neuronal function. Neuron 2009, 61(6):820-838.
4. Payne JA: Functional characterization of the neuronal-specific K-Cl
cotransporter: implications for [K+] oregulation. American Journal of
Physiology-Cell Physiology 1997, 273(5):C1516-C1525.
5. Khalilov I, Dzhala V, Ben-Ari Y, Khazipov R: Dual role of GABA in the
neonatal rat hippocampus. Developmental Neuroscience 1999, 21(3-
5):310-319.
6. Jedlicka P, Deller T, Gutkin BS, Backus KH: Activity-dependent intracellular
chloride accumulation and diffusion controls GABA-A receptor-mediated
synaptic transmission. Hippocampus 2011, 21(8):885-898.
P6
Role of topology in the spontaneous cortical activity
Silvia Scarpetta1*, Antonio de Candia2,3, Ilenia Apicella1
1
Department of Physics E.R.Caianiello & INFN, University of Salerno,
Fisciano,84084, Italy; 2Department of Physics, University of Napoli Federico
II, Napoli, Italy; 3INFN Sezione di Napoli, Napoli, Italy
E-mail: sscarpetta@unisa.it
BMC Neuroscience 2015, 16(Suppl 1):P6
Spontaneous cortical activity can show very complex collective emerging
features, with, in some cases, the alternation of down states of network
quiescence and up states of generalized spiking and neuronal
depolarization [1]. Results on in vitro and in vivo up states has suggested
that this spontaneous activity occurs in a highly structured way, with
repeating spatiotemporal patterns of cellular activity [1,2]. Because of
their stereotyped spatio-temporal dynamics, it has been conjectured that
network up states are circuit attractors that could implement memory
states [1]. The precise mechanisms by which these up states transitions
occur is still unclear, but it seems to rely on network mechanisms [1].
Previous papers [3,4] have studied a leaky IF model whose connectivity
was designed in such a manner as to favor the spontaneous emergence
of collective oscillatory spatio-temporal patterns of spikes. In this model
the alternation of up and down states does not depend on a kind of
neuron bistability, nor on synaptic depression, but is rather a network
effect. The model consists in N leaky IF units, with Poissonian noise.
Structured connections Jij was fixed with a learning procedure [3,4]
inspired to spike time dependent plasticity (STDP). After the learning and
pruning procedure [3], about 12% of the N(N-1) connections survive as
positive connections, and about 27% as negative connections. Theres a
balance condition, such that the sum of connections entering on a single
neuron jJij is of order 1/N, and therefore vanishes in the
thermodynamic limit. In order to check if the transition region with
bimodal distribution of the firing rate survive even after changes in the
topology of the network, in this work we reshuffle a fraction of the
connections. For each chosen connection we change the presynaptic
neuron, choosing as the new presynaptic neuron a random neuron of the
network. After reshuffling a fraction f of connections, not only the
number of excitatory and inhibitory connections is the same as before,
but also the strengths of the connections are the same, and only the
topology (number of loops, mean path length, clustering coefficient, etc.)
is changed. We observe that reshuffling the connections changes the
features of the dynamics dramatically. Before reshuffling, i.e. with f = 0
and connections dictated by the learning rule of [3], the system has a
transition from a regime of Poissonian noise activity to a regime of
spontaneous persistent collective replay, and at the transition point the
network dynamics shows an intermittent reactivation of the stored
patterns, with alternation of up and down state, and bimodal distribution
of spiking rate (fig. 1A). Reshuffling a small fraction of connections,
leaving all the other parameters unchanged, we observe that the
transition region with bimodal distribution disappears, and the dynamics
is Poissonian with unimodal rate distribution for all the parameters
investigated. These results shows the role of topology in dictating the
emerging collective dynamics of neural circuits.
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 19 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

Figure 1(abstract P5) A raster plot of pyramidal cell population firing; B raster plot of interneuron firing; C spatial distribution of extracellular
potassium; D intracellular chloride distribution

Figure 1(abstract P6) Distribution of the spiking rates for the model with (A) no reshuffling f = 0, (B) reshuffling a fraction f = 0.1 of
connections, and (C) all connections reshuffled f = 1. All other parameters are the same (N = 3000 neurons, H0/ = 0.221, noise variance a = 0.06 ms
1
) in the three cases
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
References
1. Cossart R, Aronov D, Yuste R: Attractor dynamics of network UP states in
the neocortex. Nature 2003, 423(6937):283-288.
2. Luczak A, MacLean JN: Default activity patterns at the neocortical
microcircuit level. Front Integr Neurosci 2012, 6:30.
3. Scarpetta de Candia: Alternation of up and down states at a dynamical
phase-transition of a neural network with spatiotemporal attractors.
Front Syst Neurosci 2014, 8:88.
4. Scarpetta S, de Candia A: Neural avalanches at the critical point between
replay and non-replay of spatiotemporal patterns. PLoS One 2013, 8(6):
e64162.
P7
A network model of neural activity in essential tremor
Nada Yousif1*, Michael Mace2, Nicola Pavese1, Roman Borisyuk3,
Dipankar Nandi1, Peter Bain1
1
Division of Brain Sciences, Imperial College London, London, W6 8RF, UK;
2
Department of Bioengineering, Imperial College London, London, SW7 2AF,
UK; 3School of Computing and Mathematics, University of Plymouth,
Plymouth, PL4 8AA, UK
E-mail: n.yousif@imperial.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P7
Deep brain stimulation (DBS) is a surgical treatment used for a number of
movement disorders, involving the chronic implantation of electrodes into
disorder specific regions in the brain. For essential tremor (ET) the targeted
brain structure is the ventralis intermedius (ViM) nucleus of the thalamus.
ET is a common movement disorder, which affects as many as 4 out of 100
adults over 40 years of age. While the cause of this disorder is unknown,
DBS works well, achieving up to 90% improvement in symptoms [1,2].
However, the mechanisms by which the therapeutic effect is obtained are
not fully understood, which in turn slows the process of finding the
optimal parameter settings which suppress symptoms and minimise
Figure 1(abstract P7) The structure of the model is shown schematically (A), and consists of the cortex, the ViM nucleus and the reticular
nucleus (nRT) of the thalamus, and the deep cerebellar nuclei (DCN). The baseline oscillatory activity of the thalamic population is within the ET
frequency range (B)
Page 20 of 200
unwanted side effects. DBS additionally provides an opportunity to record
the pathological neural activity via intraoperative recordings from the
implanted electrodes in the form of local field potentials (LFP) whilst
simultaneously recording muscle activity from the affected limbs (EMG).
Here, we present data which shows peaks in the EMG-LFP cross spectra
within the tremor frequency band (4-12 Hz). To understand the effects of a
DBS input on such pathological neural activity, we adopt a computational
modelling approach using a population representation of the network
hypothesised to underlie ET, namely cortex, cerebellum and thalamus
(Figure 1 A) and is based on previous descriptions of the essential tremor
network [3]. The model is implemented using the Wilson-Cowan approach
[4], and was simulated by exploring the parameter space to uncover
regions which produced oscillatory thalamic activity and we found that the
network exhibited oscillatory behaviour within the tremor frequency range
(Figure 1B). By applying an input to the thalamus simulating the effect of
DBS (e.g. 150Hz square pulse), we found that these oscillations are
suppressed. Therefore, this study shows that the dynamics of the ET
network are able to support oscillations at the tremor frequency.
Furthermore, the application of a DBS-like input to such a network disrupts
synchronous activity, which could explain one mechanism by which DBS
achieves therapeutic benefit.
Acknowledgements: NY is funded by a medical research grant from the
BUPA Foundation
References
1. Baizabal-Carvallo JF, Kagnoff MN, Jimenez-Shahed J, Fekete R, Jankovic J:
The safety and efficacy of thalamic deep brain stimulation in essential
tremor: 10 years and beyond. J Neurol Neurosurg Psychiatry 2014,
85:567-572.
2. Deuschl G, Raethjen J, Hellriegel H, Elble R: Treatment of patients with
essential tremor. Lancet Neurol 2011, 10(2):148-161.
3. Raethjen J, Deuschl G: The oscillating central network of Essential tremor.
Clin Neurophysiol 2012, 123(1):61-64.
4. Wilson HR, Cowan JD: Excitatory and inhibitory interactions in localized
populations of model neurons. Biophys J 1972, 12(1):1-24.
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 21 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

8. Wimmer K, Nykamp DQ, Constantinidis C, Compte A: Bump attractor

P8 dynamics in prefrontal cortex explains behavioral precision in spatial
A biophysical neural network model for visual working memory that working memory. Nat Neurosci 2014, 17:431-439.
accounts for memory binding errors
Joo Barbosa*, Albert Compte
Intitut dInvestigacions Biomdiques August Pi i Sunyer (IDIBAPS), 08036, P9
Barcelona, Spain Postsynaptic mechanisms influencing the duration of depolarization
E-mail: palerma@gmail.com discharges in hyperexcitable neuro-glial networks
BMC Neuroscience 2015, 16(Suppl 1):P8 Vasily Grigorovsky*, Berj L Bardakjian
Institute of Biomaterials and Biomedical Engineering, University of Toronto,
Binding errors, also called swap errors, occur in working memory tasks Toronto, Ontario, M5S 3G9, Canada
when the participant fails to report the feature of a previously presented E-mail: vasily.grigorovsky@utoronto.ca
target but the response is accurate relative to a non-target stimulus (for BMC Neuroscience 2015, 16(Suppl 1):P9
instance, s/he reports the location of the green item when the red items
location was to be reported). Psychophysical experiments have shown Hyperexcitability in the neural networks is one of the hallmark features of
the importance of these errors in working memory tasks [1-3], and they the epileptic brain and can manifest itself as recurrent short or long duration
reflect the failure of the system to maintain bundled through memory discharges [1]. In rodent models of Rett syndrome, it was found that long
the conjunction of features that define one object. How the brain is able duration discharges were indicative of more severe outcomes which can be
to integrate and maintain integrated several features of an item in one used as biomarkers for seizure-like activity [2]. Hyperexcitable networks
single memory is still an open question. Conjunctive selectivity, where show an increased expression of potassium afterhyperpolarization (AHP)
single units are selective to combination of stimuli, has been suggested channels [3] and persistent sodium (NaP) channels [4]. Additionally,
as a possible solution [4]. However, with the increase of features to astrocytes have been implicated in moderating neuronal firing patterns,
encode the number of necessary neurons escalates exponentially. exhibiting impaired extracellular potassium clearance in epileptic models
A hypothesis formulated for this problem in visual perception states that and potentially contributing to the development of depolarization
synchrony of different neural assemblies coding each of them a single discharges. However the precise mechanism that distinguishes
feature of an item plays the main role [5]. This is an attractive hypothesis, depolarization discharges into short and long duration has yet to be fully
because of its parsimonious and flexible neural representation: a ascertained.
particular neuron can participate in the coding of different objects at In this study, we developed a branched computational model of CA3
different times. Because under this hypothesis there is no hardwired region of hippocampus, consisting of a network of an astrocyte and a
selectivity for each group of features, it allows ex novo representations to pyramidal cell with a feedback inhibitory interneuron. As both potassium
emerge. Current neurophysiological data provides partial support for this and calcium ions have been shown to potentially affect neuronal
idea but the specific neural mechanism remains unknown [5,6]. hyperexcitability, the astrocytic model has both mechanisms - the
In order to understand this phenomenon and to test this synchrony clearance of potassium through potassium channels, and the influence of
astrocyte in the synapse.
hypothesis, we built a biophysical neural network model for the storage of
items orks for working memory (ring models, as in [7]), one representing
colors and the other one locations of equal eccentricity. These two
networks are then connected via weak cortico-cortical excitation through
AMPA synapses. With this model we are able to maintain persistent activity
through recurrent synaptic interactions in memory bumps [8] that encode
colors and locations. AMPA synapses in recurrent excitation within each
network induce gamma oscillations during bump activity through the
interplay of fast excitation and slower feedback inhibition [7]. We found
that when multiple bumps are maintained within one ring model, the
corresponding bump attractors oscillate out of synchrony with each other.
In this model, the binding between color and location is effectively
accomplished through the synchronization of gamma oscillations between
individual bumps in the two networks as a result of weak cortico-cortical
excitation. As previously proposed [9-10], in our network model different
memories are held at different phases of a low-frequency oscillation. In
some simulations swap errors arise: color bumps abruptly change their
phase and thus change their synchrony relationship with location bumps.
This leads to a wrong association between color and location that may
underlie memory binding errors reported in working memory tasks.
References
1. Wolfe JM, Cave KR: The Psychophysical Evidence for a Binding Problem
in Human Vision. Neuron 1999, 24(1):11-17.
2. Bays PM, Catalao RFG, Husain M: The precision of visual working memory
is set by allocation of a shared resource. J Vis 2009, 9(10):7.1-7.11.
3. Ma WJ, Husain M, Bays PM: Changing concepts of working memory. Nat
Neurosci 2014, 17(3):347-356.
4. Matthey L, Bays PM, Dayan P: A Probabilistic Palimpsest Model of Visual
Short-term Memory. 2015, 11(1):e1004003.
5. Singer W: Neuronal Synchrony: A versatile Code for the Definition of
Relations? Neuron 1999, 24(1):49-65.
6. Hirabayashi T, Miyashita Y: Dynamically modulated spike correlation in
monkey inferior temporal cortex depending on the feature
configuration within a whole object. J Neurosci 2005, Figure 1(abstract P9) Effects of AHP channel blockade on the
25(44):10299-10307. duration of epileptic discharges. Somatic voltage under A) normal
7. Compte A, Brunel N, Goldman-Rakic PS, Wang XJ: Synaptic mechanisms channel expression, B) partial dendritic AHP blockade, and C) full
and network dynamics underlying spatial working memory in a cortical dendritic AHP blockade
network model. Cereb Cortex 2000, 10(9):910-923.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Preliminary results show that in hyperexcitable systems with fully working
potassium AHP channels depolarization discharges cease after less than a
second (Fig. 1A), classifying them as short duration. With partial dendritic
AHP channel blockade the duration of discharges increases, transitioning
into long duration discharges (Fig. 1B). In the extreme case where all of the
dendritic AHP channels are blocked, there is no cessation of the seizure-like
state (Fig. 1C). We hypothesize that in hyperexcitable systems the
postsynaptic AHP (and to a lesser extent the Nap) channels work in concert
with glial cells to control the duration of depolarization discharges.
Acknowledgements: We would like to thank CIHR and NSERC for
providing the opportunity to investigate neuron-glia interactions
References
1. Lang M, Wither RG, Colic S, Wu C, Monnier PP, Bardakjian BL, Zhang L,
Eubanks JH: Rescue of behavioral and EEG deficits in male and female
Mecp2-deficient mice by delayed Mecp2 gene reactivation. Human
Molecular Genetics 2014, 23(2):303-318.
2. Colic S, Min Lang, Wither RG, Zhang Liang, Eubanks JH, Bardakjian BL:
Characterization of HFOs in short and long duration discharges
recorded from in-vivo MeCP2-deficient mice. Conf Proc IEEE Eng Med Biol
Soc 2014, 4603-4606.
3. Schulz R, Kirschstein T, Brehme H, Porath K, Mikkat U, Khling R: Network
excitability in a model of chronic temporal lobe epilepsy critically
depends on SK channel-mediated AHP currents. Neurobiol Dis 2012,
45(1):337-347.
4. Estacion M, Gasser A, Dib-Hajj SD, Waxman SG: A sodium channel
mutation linked to epilepsy increases ramp and persistent current of
Nav1.3 and induces hyperexcitability in hippocampal neurons. ExpNeurol
2010, 224(2):362-368.
P10
Different weightings of input components to hippocampal CA1 place
cells in young and aged rats
Frances S Chance1*, Andrew P Maurer2, Sara N Burke2, Carol A Barnes3,4,5
1
Department of Data Driven and Neural Computing, Sandia National
Laboratories, Albuquerque, NM 87123, USA; 2Department of Neuroscience,
University of Florida, Gainesville FL 32611, USA; 3Evelyn F. McKnight Brain
Institute, University of Arizona, Tucson, AZ 85721, USA; 4ARL Div. of Neural
Systems, Memory & Aging, University of Arizona, Tucson, AZ 85721, USA;
5
Departments of Psychology, Neurology and Neuroscience, University of
Arizona, Tucson, AZ 85721, USA
E-mail: fschanc@sandia.gov
BMC Neuroscience 2015, 16(Suppl 1):P10
Hippocampal place cells vary both their firing rate and also the timing of
their action potentials relative to the theta rhythm as an animal moves
through space, suggesting that these neurons utilize both these features
to encode spatial information. Place cells within the CA1 subfield receive
synaptic input from CA3 (via the Schaffer collateral pathway) and also EC3
(layer 3 of entorhinal cortex via the perforant pathway). These pathways
are thought to carry different modalities of information, for example auto-
associative memories from CA3 and more sensory-driven information from
entorhinal cortex. We seek to understand how both modalities of
information are represented in the spiking outputs of CA1 place cells.
One model of CA1 phase precession [1] has proposed that these two input
components drive CA1 spiking over spatially-offset but overlapping ranges
of positions, and also over different ranges of theta phases. Because CA3
and EC3 inputs drive spiking at different theta phases in this model,
changes in relative input strengths can be quantified by examining spike
theta-phase distributions. For example, when one input component
dominates CA1 spiking, the result is a unimodal theta-phase distribution,
but if both input components drive CA1 spiking, their different theta-
phase ranges produce a more bimodal theta-phase distribution.
In this study, we examine the spike patterns of CA1 place cells recorded
from young and aged rats. Aged rat brains show a decreased number of
functional synapses from CA3 to CA1 pyramidal cells compared to brains
from younger rats (for review, see [2,3]). This suggests that theta-phase
distributions of CA1 place-cell spiking will change with aging. We find that
place cells from aged rats are more strongly theta-modulated, suggesting
that CA3 input plays an important role in either driving or facilitating a
second spiking component of CA1 activity.
Page 22 of 200
Acknowledgements: Supported by the McKnight Brain Research
Foundation, NIH Grant AG012609, and NIH Grant NS070464. Sandia National
Laboratories is a multi-program laboratory managed and operated by
Sandia Corporation, a wholly owned subsidiary of Lockheed Martin
Corporation, for the U.S. Department of Energys National Nuclear Security
Administration under contract DE-AC04-94AL85000. (SAND2015-1198A).
References
1. Chance FS: Hippocampal phase precession from dual input components.
J Neurosci 2012, 32(47):16693-16703.
2. Rosenzweig ES, Barnes CA: Impact of aging on hippocampal function:
plasticity, network dynamics, and cognition. Prog Neurobiol 2003,
69(3):143-179.
3. Burke SN, Barnes CA: Senescent synapses and hippocampal circuit
dynamics. Trends Neurosci 2010, 33(3):153-161.
P11
Mechanisms of hippocampal sequence replay
Paola Malerba*, Giri P Krishnan, Maxim Bazhenov
Cell Biology and Neuroscience, University of California Riverside, Riverside,
CA 92507, USA
E-mail: paola.malerba@ucr.edu
BMC Neuroscience 2015, 16(Suppl 1):P11
Sleep is known to be important for memory consolidation [1], and memories
are thought to be stored in the hippocampus during wakefulness and
transferred to cortex during sleep [2]. Recently, memory replay - repeatable
sequences of pyramidal cell firing - has been demonstrated during sleep, and
associated with characteristic brain oscillations, giving rise to the hypothesis
that these may form the critical neural substrate of memory consolidation. It
is known that the content of hippocampal replay can be biased during sleep,
in what is called cued-reactivation [3], and that similar paradigms show
enhanced memory performance in humans [4]. Moreover, tampering with
replay can disrupt memory formation and consolidation [5]. Despite
extensive evidence highlighting the importance of replay within the broader
phenomenon of sleep-mediated memory consolidation, the mechanisms
underlying sequence replay are still unknown.
During sleep, replay events are associated with specific patterns of
neuronal oscillations [6]. Replay is seen in cortex during slow oscillation -
a rhythmic (< 1Hz) state in which periods of activity (active or Up states)
alternate with quiet periods (silent or Down states), while replay in the
hippocampus is associated with sharp-wave ripple events - irregularly
brief bouts of high frequency (>150 Hz) firing, driven by strong excitatory
inputs coming from CA3, which result in a strong deflection in the LFP of
stratum radiatum (the sharp wave) [7].
In the present study, we build on our previous research [8] to develop a
model of spike sequence replay during sleep. In the past, we have
introduced a model of CA1 ripples in which oscillations are transients,
mediated by the intrinsic frequency of CA1 basket cells driven by CA3
activation. In this work, we construct a model of CA3 area in which
stochastic intrinsic activation of pyramidal cells originates a massive cell
activation that results in a strong excitatory input to area CA1. We observe
that sequential activation of a selected sub-group of CA1 pyramidal cells is
driven by a less specific sequential activation in CA3. We then characterize
the mechanisms underlying sequence selection and reactivation within a
sharp-wave ripple event.
Using hippocampal and thalamocortical models, we investigate the role
of cortical inputs on ripple timing and their specific spike content. Our
study illustrates the possible role of cortical Up states during slow
oscillation in biasing hippocampal replay, which is a core component of
the cortico-hippocampal interaction underlying memory consolidation.
Acknowledgements: Supported by ONR (MURI: N000141310672).
References
1. Rasch B, Born J: About sleeps role in memory. Physiol Rev 2013, 93(2):681-766.
2. Marr D: Simple memory: a theory for archicortex. Philos Trans R Soc Lond
B Biol Sci 1971, 262(841):23-81.
3. Bendor D, Wilson MA: Biasing the content of hippocampal replay during
sleep. Nat Neurosci 2012, 15(10):1439-1444.
4. Cairney SA, Durrant SJ, Hulleman J, Lewis PA: Targeted memory
reactivation during slow wave sleep facilitates emotional memory
consolidation. Sleep 2014, 37(4):701-707, 707A.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
5. Nakashiba T, Buhl DL, McHugh TJ, Tonegawa S: Hippocampal CA3 output
is crucial for ripple-associated reactivation and consolidation of memory.
Neuron 2009, 62(6):781-787.
6. Ji D, Wilson MA: Coordinated memory replay in the visual cortex and
hippocampus during sleep. Nat Neurosci 2007, 10(1):100-107.
7. Ylinen A, Bragin A, Nadasdy Z, Jando G, Szabo I, Sik A, Buzsaki G: Sharp
wave-associated high-frequency oscillation (200 Hz) in the intact
hippocampus: network and intracellular mechanisms. J Neurosci 1995,
15(1 Pt 1):30-46.
8. Bazhenov M, Timofeev I, Steriade M, Sejnowski TJ: Model of
thalamocortical slow-wave sleep oscillations and transitions to activated
States. J Neurosci 2002, 22(19):8691-8704.
P12
Bayesian supervised learning and state estimation in a model of the
cerebellum
Benjamin Campbell
Laboratory of Biological Modeling, The Rockefeller University, New York, NY,
10065, USA
E-mail: neuro@bccampbell.ca
BMC Neuroscience 2015, 16(Suppl 1):P12
The last decade has seen an explosion of research on the topic of Bayesian
inference in neural systems. The cerebral cortex is now widely believed to
form predictive internals models of the environment. But what does this
mean for the cerebellum, which has historically been attributed just this
function? And what might Bayesian inference mean in the cerebellar
context?
I present a cerebellar model unifying the Bayesian framework with classical
cerebellar theories of supervised learning, state estimation, and control
theory. This model builds on the dominant supervised learning theory of
cerebellar function, which traces to the pioneering work of Marr. This
model is extended to account for Bayesian optimal updating of the
uncertainties inherent in probabilistic inference. Further, the model is
capable of learning about states with little to no supervised input. It also
provides theoretical explanation for a wide range of phenomenology
observed of the circuit most importantly unsupervised synaptic plasticity
upstream of cerebellar Purkinje cells.
The connectionist architecture of this model is closely based on that of the
cerebellum. In the input stage, unsupervised learning in the cerebellar
granular layer is seen as essential for subsequent supervised learning in
the Purkinje layer. Based on Golgi cell feedback, the granular layer learns
to encode supervised learning covariances in a manner akin to the kernel
learning of Gaussian processes. This circuit then predicts and filters mossy
fiber input, conveying only select signals downstream for supervised
learning. This granular representation is further influenced by a global
serotonergic signal conveyed by Lugaro cells. In the cerebellar molecular
layer, inhibitory interneurons then provide higher-order corrections to
these estimates. Finally, learning in the cerebellar nuclei is seen as a
parallel, simpler, regression model, allowing for faster, courser adjustments,
corrected in time by the cerebellar cortical circuit.
P13
Investigating intrinsic and evoked activities in cultured neuronal
networks by dimensional reduction techniques and high-density MEAs
Thierry Nieus*, Stefano Di Marco, Alessandro Maccione, Hayder Amin,
Luca Berdondini
Neuroscience Brain Technology Istituto Italiano di Tecnologia, via Morego 30,
16163 Genoa, Italy
E-mail: thierry.nieus@iit.it
BMC Neuroscience 2015, 16(Suppl 1):P13
High density microelectrode arrays (MEAs) provide extracellular recordings
from thousands of closely spaced electrodes and with sub-millisecond
resolution. These devices offer thus novel capabilities to investigate the
interplay between ongoing and evoked electrophysiological signaling
within networks in-vitro. However, to effectively take advantage from the
spatiotemporal resolution of these MEAs, adapted analysis tools are
needed. Here we report on our recent advancements toward this goal.
A novel high density MEA with on-chip stimulating electrodes was used
Page 23 of 200
to record from 4096 electrodes and electrical stimulation was delivered
alternatively through one of the 16 equally spaced electrodes on
hippocampal primary cultures derived from mouse. The evoked activities
propagated reliably across the network and were specific to the
stimulating electrode [1]. Moreover, from the second week in vitro cell
cultures also displayed synchronous like events (called synchronous burst
events, SBEs) that propagated similarly to the evoked activities across the
entire network. These propagations might be informative of the
underlying network connectivity and their classification based on the
spatiotemporal patterns might elucidate the networks organization and
its ongoing dynamic. Former attempts, in classifying SBEs, considered
simplified descriptors of neural activity (e.g. center activity trajectory [2]).
Here, we have adopted a more rigorous approach by applying dimensional
reduction techniques (PCA) that take advantage of the redundancy and of
the sparseness of multi-unit recordings. We found that a large fraction (i.e.
>50%) of the variance of the network events (either evoked or
spontaneous) was explained by as few as 3 principal components (PCs). By
increasing the number of PCs (i.e. ~10) up to 80% of the total variance of
the network events could be explained. Thus, the PCA approach
constitutes an effective methodology to represent the spontaneous/
evoked events in a lower dimensional space. As a consequence the
principled PCA methodology we developed improved the clustering of the
network events respect to existing methodologies [2] in different
experimental settings (e.g. treatment with chemical compounds). Finally,
the PCA clustering approach also allowed to infer on state dependent
processing phenomena occurring in these networks.
Acknowledgements: We acknowledge the financial support of the SI-
CODE project of the Future and Emerging Technologies (FET) programme
within the Seventh Framework Programme for Research of The European
Commission, under FET-Open grant number: FP7-284553.
References
1. Maccione A, Nieus T, Simi A, Amin H, Gandolfo M, Berdondini L: Multi-site
electrical stimulation integrated on 4096 high density micro electrode arrays
(MEAs) reveals the effective connectivity of dissociated neuronal cultures SFN
2012, October 13th-17th, New Orleans, USA.
2. Gandolfo M, Maccione A, Tedesco M, Martinoia S, Berdondini L: Tracking
burst patterns in hippocampal cultures with high-density CMOS-MEAs.
J Neural Eng 2010, 7(5):056001.
P14
Cell assembly dynamics of sparse inhibitory networks: a simple model
for the activity of the Medium Spiny Neurons
David Angulo-Garcia1*, Alessandro Torcini1, Joshua D Berke2
1
Istituto dei Sistemi Complessi, Consiglio Nazionale delle Ricerche (CNR), via
Madonna del Piano 10, Sesto Fiorentino, Italy I-50019; 2Department of
Psychology, University of Michigan, Ann Arbor, 530 Church St., Ann Arbor, MI
48104, USA
E-mail: david.angulo@fi.isc.cnr.it
BMC Neuroscience 2015, 16(Suppl 1):P14
Here we show that a simple inhibitory network model, made of sparsely
connected Leaky Integrate and Fire (LIF) neurons, is able to retrieve some
of the relevant dynamical features of a striatal network, in particular the
appearance of cell assembly dynamics as it has been reported in in-vitro
experiments of rats striatum [1]. In a first approach, we discuss how our
simple model is consistent with the findings in [2,3]. For an optimal choice
of the model parameters, the response of the neurons to uniformly
distributed constant inputs, occurs in a bursting fashion. As shown in
Figure. 1B, the neurons organize their dynamics in groups with correlated
bursting-like activity, displaying typical recurrent patterns, similarly to the
dynamics of Medium Spiny Neurons (MSNs). Furthermore, the firing of the
neurons taking part in this structured cell assembly dynamics is
characterized by a high variability, as shown in Figure 1A for a few
representative neurons. This high variability is reflected in a coefficient of
variation of the interspike-interval (ISI) larger than one. An important
aspect of the dynamics of the MSNs is the emergence of coexisting
correlated and anti-correlated assemblies, as reported in the experimental
work by Carrillo-Reid et al. [1]. Indeed also in our system this aspect is
present, as revealed by examining the cross-correlation matrix of the firing
rates shown in Figure 1C. Here the neurons are grouped in assemblies
accordingly to their level of correlation (as in Figure 1B) and it is evident
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P14) A. Firing rates of 5 selected neurons; B. raster plot activity, the firing times are colored accordingly to the assembly the
neurons belong to; C. cross-correlation matrix of the firing rates. The neurons in panel B and C are clusterized accordingly to the correlation of their firing
rates, by employing the k-means algorithm
that the correlated activities within the neuronal assemblies can be highly
anti-correlated with other cells in the network.
Acknowledgements: This work has been supported by the European
Commission under the program Marie Curie Network for Initial Training,
through Project No. 289146, Neural Engineering Transformative Technologies
(NETT). D.A.-G. also acknowledges the partial support provided by
Departamento Administrativo de Ciencia Tecnologia e Innovacion-
Colciencias through the program Doctorados en el exterior-2013.
References
1. Carrillo-Reid L, Tecuapetla F, Tapia D, Hernndez-Cruz A, Galarraga E,
Drucker-Colin R, Bargas J: Encoding network states by striatal cell
assemblies. Journal of Neurophysiology 2008, 99(3):1435-1450.
2. Ponzi A, Wickens J: Sequentially switching cell assemblies in random
inhibitory networks of spiking neurons in the striatum. The Journal of
Neuroscience 2010, 30(17):5894-5911.
3. Ponzi A, Wickens JR: Optimal balance of the striatal medium spiny
neuron network. PloS Computational Biology 2013, 9(4):e1002954.
P15
Sharp wave-ripple complexes in a reduced model of the hippocampal
CA3-CA1 network of the macaque monkey
Juan F Ramirez-Villegas1,2*, Nikos K Logothetis1,3, Michel Besserve1,4
1
Department of Physiology of Cognitive Processes, Max Planck Institute for
Biological Cybernetics, Tbingen, 72076, Germany; 2Graduate School of
Neural & Behavioral Sciences, International Max Planck Research School,
Eberhard-Karls University of Tbingen, Tbingen, 72074, Germany; 3Centre
for Imaging Sciences, Biomedical Imaging Institute, The University of
Manchester, Manchester, M13 9PT, UK; 4Department of Empirical Inference,
Max Planck Institute for Intelligent Systems, Tbingen, 72076, Germany
E-mail: juan.ramirez-villegas@tuebingen.mpg.de
BMC Neuroscience 2015, 16(Suppl 1):P15
Sharp wave-ripple complexes observed in the hippocampal CA1 local field
potential (LFP) are thought to play a major role in memory reactivation,
transfer and consolidation. SPW-Rs are known to result from a complex
interplay between local and upstream hippocampal ensembles. However,
the key mechanisms that underlie these events remain partly unknown. In
this work, we introduce a reduced, but realistic multi-compartmental model
of the macaque monkeys hippocampal CA3-CA1 network. The model
consists of two semi-linear layers, each consisting of two-compartmental
pyramidal neurons and one-compartmental perisomatic-targeting basket
cells. Connections in the network were modeled as AMPA synapses, based
on physiological and anatomical data. Notably, while auto-association fibers
were prevalent in CA3, CA1 connectivity -inspired by recent findings-
implemented a feedback and reciprocal inhibition, dominated by recurrent
inhibition and pyramidal cells-interneurons synapses. SPW-R episodes
emerge spontaneously in the CA1 subfield LFP (which is assumed
proportional to transmembrane currents across all compartments and
medium resistivity): Episodes of short-lived high-frequency oscillations
(ripples, 80-180 Hz) on top of a massive dendritic depolarization (< 20 Hz)
with visual and quantitative characteristics observed experimentally [1].
Page 24 of 200
Concomitantly, the CA3 subfield LFP presents episodes of quasi-
synchronous neuronal bursting in the form of gamma episodes (25-75 Hz).
The model reveals a lower bound for the minimal network that may
generate SPW-R activity, and predicts a large number of features of in vivo
hippocampal recordings in macaque monkeys [1]. Spike-LFP coherence
analysis in CA1 displays reliable synchrony of spiking activity in the ripple
LFP frequency band, suggesting that modeled SPW-R episodes reflect a
genuine network oscillatory regime. Interestingly, interneuronal firing shows
coherence increases concomitant with the beginning and the end of the
SPW-R event, together with increases over gamma frequencies.
The model suggests that activity of both pyramidal neurons and
interneurons is critical for the local genesis and dynamics of physiological
SPW-R activity. Unlike other models, we found that it is interneuronal
silence, not interneuronal firing that triggers these fast oscillatory events, in
line with the fact that unbalanced excitability of selected pyramidal cells
marks the beginning of single network episodes. Interneuronal silence
quickly increases population firing of pyramidal cells. The interneuronal
population activity increases with some latency due to the unbalanced
excitatory drive, becoming pivotal to pyramidal cell activity, and further
pacing pyramidal cells due to interneuronal fast kinetic properties. Our
modeled data suggests that this effect is possibly mediated by a silencing-
and-rebound-excitation mechanism, maintaining the frequency of the field
oscillation bounded to the ripple range. The reduced model suggests a
simple mechanism for the occurrence of SPW-Rs, in light of recent
experimental evidence. We provide new insights into the dynamics of the
hippocampal CA3-CA1 network during ripples, and the relation between
neuronal circuits activity at meso- and microscopic scales. Finally, our model
exhibits characteristic cell type-specific activity that might be critical for the
emergence of physiological SPW-R activity and therefore, for the formation
of hippocampus-dependent memory representations.
Reference
1. Logothetis NK, Eschenko O, Murayama Y, Augath M, Steudel T, Evrard HC,
Besserve M, Oeltermann A: Hippocampal-cortical interaction during
periods of subcortical silence. Nature 2012, 491(7425):547-553.
P16
Modelling the mechanoreceptors dynamic behaviour
Zhuoyi Song1*, Robert W Banks2, Guy S Bewick3
1
Centre for Mathematic, Physics and Engineering in the Life Sciences and
Experimental Biology (CoMPLEX), University College London, London, UK;
2
School of Biological and Biomedical Sciences, University of Durham,
Durham, DH1 3LE, UK; 3School of Medical Sciences, Institute of Medical
Sciences, University of Aberdeen, Aberdeen, AB25 2ZD, UK
E-mail: zhuoyi.song@ucl.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P16
All sensory receptors adapt, i.e., they constantly adjust their sensitivity to
external stimuli to match the current natural environment [1].
Electrophysiological responses of sensory receptors from widely different
modalities seem to exhibit common features related to adaptation, and
these features can be used to examine the underlying sensory transduction
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
mechanisms [1,2]. Among the principal senses, mechanosensation remains
the least understood at the cellular level [3]. To gain greater insights into
mechanosensory signalling, we investigated if mechanosensation displayed
adaptive dynamics that could be explained by similar biophysical
mechanisms in other sensory modalities. To do this, we adapted a fly
photoreceptor model [4] to describe the primary transduction process for a
stretch-sensitive mechanoreceptor, taking into account the viscoelastic
properties of the accessory muscle fibres [5] and the biophysical properties
of known mechanosensitive channels (MSCs). The models output is in
remarkable agreement with the electrical properties of a primary ending of
an isolated decapsulated spindle; ramp-and-hold stretch evokes a
characteristic pattern of potential change, consisting of a large dynamic
depolarization during the ramp phase and a smaller static depolarization
during the hold phase [6]. The initial dynamic component is likely to be
caused by both the mechanical properties of the muscle fibres and a
refractory state of MSCs. Consistent with literature, the current model
predicts that the dynamical component is due to a rapid stress increase
during the ramp [7]. More novel predictions from the model are the
mechanisms to explain the initial peak in the dynamical component. At the
onset of the ramp, all MSCs are sensitive to external stimuli, but as they
become refractory (clipped inactivated state), fewer MSCs are able to
respond to the continuous stretch, causing a sharp decrease after the peak
response. The same mechanism could contribute a faster component in
sensory habituation of a mechanoreceptor, in which a receptor responds
more strongly to the first stimulus episode during repetitive stimulation [8].
References
1. De Palo G, Facchetti G, Mazzolini M, Menini A, Torre V, Altafini C: Common
dynamical features of sensory adaptation in photoreceptors and
olfactory sensory neurons. Scientific Reports 2013, 3:1251.
2. Torre V, Ashmore JF, Lamb TD, Menini A: Transduction and adaptation in
sensory receptor cells. J Neurosci 1995, 15(12):7757-7768.
3. Chalfie M: Neurosensory mechanotransduction. Nat Rev Mol Cell Bio 2009,
10:44-52.
4. Song ZY, Postma M, Billings SA, Coca D, Hardie RC, Juusola M: Stochastic,
Adaptive Sampling of Information by Microvilli in Fly Photoreceptors.
Curr Biol 2012, 22(15):1371-1380.
5. Swerup C, Rydqvist B: A mathematical model of the crustacean stretch
receptor neuron. J Neurophysiol 1996, 76(4):2211-2220.
6. Hunt CC, Wilkinson RS, Fukami Y: Ionic basis of the receptor potential in
primary endings of mammalian muscle spindles. J Gen Physiol 1978,
71(6):683-698.
7. Matthews PB: Muscle Spindles and Their Motor Control. Physiol Rev 1964,
44(2):219-288.
8. Pasztor VM, Bush BMH: Graded Potentials and Spiking in Single Units of
the Oval Organ, a Mechanoreceptor in the Lobster Ventilatory System
.3. Sensory Habituation to Repetitive Stimulation. J Exp Biol 1983,
107:465-472.
P17
The functional significance of fasciculation and repulsion in a
computational model of axon growth
Robert Merrison-Hort1*, Oliver Davis2, Roman Borisyuk1
1
School of Computing and Mathematics, Plymouth University, Plymouth,
Devon, PL4 8AA, UK; 2Brighton and Sussex Medical School, Brighton, East
Sussex, BN1 9PX, UK
E-mail: robert.merrison@plymouth.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P17
The neurons in the developing spinal cord of a Xenopus tadpole form a
simple network that is able to produce behaviors such as swimming and
struggling from an early stage. At present, however, the developmental
processes that allow such networks to form are not well understood. We
use computational models to shed light on which features of development
are important for the formation of networks that can produce patterns of
neuronal activity corresponding to a particular behavior. Previously, we
have shown that a simple model in which axon growth is guided by
sensitivities to gradients of chemical cues can produce axonal trajectories
that closely match the available anatomical data [1]. Furthermore, when
the synaptic connectivity derived from these generated axons is mapped
onto a large scale physiological model, the resulting network responds to
simulated skin touch input by generating a stable pattern of anti-phase
Page 25 of 200
rhythmic activity which resembles that seen in real tadpoles during
swimming [2], as well as reproducing other experimental observations
such as transient synchrony [3].
In this poster we demonstrate the effects of adding axon fasciculation and
repulsion mechanisms to our growth model. Axon fasciculation is a
process whereby a developing axon can detect the presence of another
nearby axon and begin to grow along the existing axon; repulsion is the
opposite process, where axons actively avoid each other. Both fasciculation
and repulsion have been observed in the growth of commissural axons in
the spinal cord of very early stage tadpoles [4], so we sought to use our
computational model to investigate the possible function of these
processes. The existing model was adjusted such that the growth angle of
each axon contained an additional term based on the influence of nearby
axons. This influence could be attractive or repulsive, and could vary based
on the current position of the axon tip, since it appears that commissural
axons change from repulsion to fasciculation after crossing the ventral
mid-line. Adding these processes to the model caused dramatic changes
to the pattern of axonal trajectories that were immediately visible (Figure
1), as well as quantitative changes to various anatomical characteristics,
such as the total number of synapses formed. This poster shows the
effects that fasciculation and repulsion have on both the generated
anatomy and the dynamical behavior of the spiking neuron model, and
discusses the possible biological and medical significance of these results
(e.g. for axon regeneration).
References
1. Borisyuk R, Azad AK, Conte D, Roberts A, Soffe SR: A developmental
approach to predicting neuronal connectivity from small biological
datasets: a gradient-based neuron growth model. PLoS One 2014, 9(2):
e89461.
2. Roberts A, Conte D, Hull M, Merrison-Hort R, Azad AK, Buhl E, Borisyuk R,
Soffe SR: Can simple rules control development of a pioneer vertebrate
neuronal network generating behavior? J Neurosci 2014, 34(2):608-621.
3. Li WC, Merrison-Hort R, Zhang H-Y, Borisyuk R: The Generation of
Antiphase Oscillations and Synchrony by a Rebound-Based Vertebrate
Central Pattern Generator. J Neurosci 2014, 34(17):6065-6077.
4. Moon MS, Gomez TM: Adjacent pioneer commissural interneuron growth
cones switch from contact avoidance to axon fasciculation after midline
crossing. Dev Biol 2005, 288(2):474-486.
P18
Are rich club regions masters or slaves of brain network dynamics?
Leonardo L Gollo1*, Andrew Zalesky2,3, R Matthew Hutchison4,
Martijn van den Heuvel5, Michael Breakspear1
1
Systems Neuroscience Group, QIMR Berghofer, Brisbane, Queensland, QLD
4006, Australia; 2Melbourne Neuropsychiatry Centre, Melbourne, Victoria,
Australia; 3Melbourne Health, Melbourne, Victoria, Australia; 4Center for Brain
Science, Harvard University, Cambridge, MA, USA; 5Brain Center Rudolf
Magnus, University Medical Center, Utrecht, the Netherlands
E-mail: leonardo.l.gollo@gmail.com
BMC Neuroscience 2015, 16(Suppl 1):P18
Anatomical networks of brain systems emphasize their inherent hierarchical
structure, in which the rich club (most inner core of the brain) is highlighted
because it promotes topological efficiency and integrates communication
between regions [1]. The hierarchical structure is fundamental and also
observed in the neuronal dynamics: Cortical anatomy is thought to
recapitulate the temporal hierarchy that is inherent in the dynamics of
environmental states and human behavior [2]. According to this classical
view the caudal-rostral arrangement maps the fast-slow gradient of
neuronal time scales [3]. Here we ask whether the topological arrangement
of the networks can provide a scaffold on which a hierarchy of time scales
including slow and stable dynamics emerges. We modeled neuronal
dynamics unfolding on an anatomical network reconstructed from primate
cortex that has a set of rich-club regions at the core of the network that are
surrounded by their feeders and peripheral areas [4]. We utilized a
conductance-based neural mass model [5], which gives rise to node
dynamics with two time scales. The system synchronizes at the fast time
scale and exhibits rich phase relations at the slow rhythm (~10Hz). We then
characterized and compared the oscillatory dynamics of nodes and the
synchronization between pairs of nodes. Our results show that highly
connected regions, such as rich-club members, exhibit a more regular
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 26 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

Figure 1(abstract P17) Example trajectories of commissural interneuron axons from the original model (A) and with fasciculation (B). There are
a small number of pioneer axons (red), which grow earlier than the main group (blue). Axons start growing from the somas (top, stars) and grow
ventrally to cross the mid-line (dorso-vental position 0m), before turning to grow in the ascending direction (rostrally). Every axon also has a secondary
branch which grows in the descending direction (pioneers: magenta, others: green)

dynamics, and larger stability of phase relation with respective to other
regions [6]. Owing to the dynamics of the local connectivity and network-
motif structure [7,8], central nodes also catalyze better local zero-lag
synchronization between their neighbors. Together our results show how
the topology of this constellation of brain regions supports stable, slowly
fluctuating patterns of synchronization. In contrast, the topology of the
surrounding feeder cortical regions shows unstable, rapidly fluctuating
dynamics. Hence, assuming the interaction of identical dynamical elements
coupled via the structural connectome of primate cortex [4], we already find
an emerging hierarchy of time scales due to the network structure. We thus
propose that the network properties of central cortical regions are the
structural determinants of slow neuronal fluctuations in these brain regions.
Although the traditional caudal-rostral hierarchical cortical mapping
exhibited an unprecedented success, our dynamic-based study aligned with
recent observations indicates that the cortical organizing principle may be
best described by a periphery-core axis. Intriguingly, as a metaphor to social
networks, rich-club members could be seen to be slaves of their own
power. The richness of these brain regions comes from their large in-degree.
However, for this very reason, their dynamics are consequently more regular
because larger in-degree guarantees larger regularity. In other words, the
autonomous dynamics of connectome hubs are largely enslaved to the
strong, rhythmic output of the entire connectome. Therefore, rich-club
regions cannot behave very differently from what is expected in contrast
to the peripheral nodes, which have the most freedom to explore the
dynamical landscape.
References
1. van den Heuvel MP, Sporns O: Network hubs in the human brain. Trends
Cogn Sci 2013, 17(12):683-696.
2. Kiebel SJ, Daunizeau J, Friston KJ: A hierarchy of time-scales and the
brain. PLoS CB 2008, 4(11):e1000209.
3. Fuster JM: The prefrontal cortex-an update: time is of the essence.
Neuron 2001, 30(2):319-333.
4. Harriger L, Van Den Heuvel MP, Sporns O: Rich club organization of
macaque cerebral cortex and its role in network communication. PloS
One 2012, 7(9):e46497.
5. Breakspear M, Terry JR, Friston KJ: Modulation of excitatory synaptic
coupling facilitates synchronization and complex dynamics in a
biophysical model of neuronal dynamics. Network 2003, 14(4):703-732.
6. Gollo LL, Zalesky A, Hutchison RM, van den Heuvel M, Breakspear M:
Dwelling quietly in the rich club: brain network determinants of slow
cortical fluctuations. Phil Trans R Soc B 2015, 370:1668.
7. Gollo LL, Mirasso C, Sporns O, Breakspear M: Mechanisms of zero-lag
synchronization in cortical motifs. PLoS Comput Biol 2014, 10:e1003548.
8. Gollo LL, Breakspear M: The frustrated brain: from dynamics on motifs to
communities and networks. Phil Trans R Soc B 2014, 369(1653):20130532.
P19
Multi-compartmental modeling in Brian 2
Marcel Stimberg1,2,3*, Dan F M Goodman4, Romain Brette1,2,3
1
Sorbonne Universits, UPMC Univ. Paris 06, UMR_S 968, Institut de la Vision,
Paris, F-75012, France; 2INSERM, U968, Paris, F-75012, France; 3CNRS,
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
UMR_7210, Paris, F-75012, France; 4Department of Electrical and Electronic
Engineering, Imperial College London, London, UK
E-mail: marcel.stimberg@inserm.fr
BMC Neuroscience 2015, 16(Suppl 1):P19
Brian 2 is a fundamental rewrite of the popular Brian [1,2] simulator for
spiking neural networks. It is written in the Python programming language
and focuses on simplicity and extensibility: neuronal and synaptic models
can be described using mathematical formulae and with the use of
physical units [3].
Brian has been traditionally used to simulate networks of single-
compartment neurons; complex morphologies, modeled as multi-
compartment neurons, were not supported. We have recently added
support for such models: a neurons morphology can be either created
from a file (in the SWC format [4]) or be created iteratively in the
simulation script itself, using a convenient syntax. The flexibility of the
Brian simulator applies to multi-compartment models in the same way as
to single-compartment models: the model dynamics are governed by
arbitrary user-provided differential equations, models can be flexibly
parameterized over all the compartments and all variables describing the
internal state of the model (voltages, currents, conductances, etc) are
accessible to the user.
The simulation of multi-compartment models is more complex than that of
non-electrically coupled single-compartment models: in addition to the
numerical integration of the differential equations governing the change
of the membrane potential in a single compartment, the currents flowing
across neighboring compartments have to be taken into account. To this
end, Brian employs a domain-decomposition method which corrects an
initial solution valid for uncoupled compartments, taking the actual
coupling into account. This method has the advantage that it is easy to
parallelize, since most of computations are performed independently for
each compartment. We demonstrate the advantage of this by parallelizing
the computation over multiple processor cores using the OpenMP
framework. For the future, we are planning to further accelerate the
simulation speed by making use of the massively-parallel processing
capabilities of graphical processing units.
With this addition, we have significantly enlarged the range of research
questions that can be investigated using Brian.
Brian is made available under a free software license and all development
takes place in public code repositories.
Acknowledgements: This work was partly supported by grant ANR-14-
CE13-0003.
References
1. Goodman DFM, Brette R: The Brian Simulator. Frontiers in Neuroscience
2009, 3(2):192-197.
2. The Brian spiking neural network simulator. [http://briansimulator.org],
accessed 21-02-2015.
3. Stimberg M, Goodman DFM, Benichoux V, Brette R: Equation-oriented
specification of neural models for simulations. Frontiers in
Neuroinformatics 2014, 8:6.
4. Turner DA, Pyapali GK, Wheal HV: An on-line archive of reconstructed
hippocampal neurons. Journal of Neuroscience Methods 1998, 84(1-2):49-54.
P20
Emergence of ITD tuning in the MSO with a realistic periphery model
P Yger1,2,3,4*, V Benichoux1,2,3,4, M Stimberg1,2,3,4, R Brette1,2,3,4
1
Institut dEtudes de la Cognition, Ecole Normale Suprieure, Paris, France;
2
Sorbonne Universits, UPMC Univ. Paris 06, UMR S 968, Institut de la Vision,
Paris, F-75012, France; 3INSERM, U968, Paris, F-75012, France; 4CNRS, UMR
7210, Paris, F-75012, France
E-mail: pierre.yger@inserm.fr
BMC Neuroscience 2015, 16(Suppl 1):P20
To localize sounds in the environment, animals mostly rely on spectro-
temporal cues originating from the physical disparities of the sound
waveforms impacting the two ears. Among those, the Interaural Time
Difference (ITD) has been shown to be crucial in mammals for locating
low-frequency sounds, and is known to be processed by neurons in a
particular structure, the Medial Superior Olive (MSO). While it is classically
considered that the emergence of ITD selectivity in a neuron of the MSO is
simply due to differences in the axonal delays originating from the two
Page 27 of 200
ears and impinging those neurons (the so called delay-line model [1]),
experimental evidence shows that the best delay (the ITD at which the
neurons firing rate is maximum) is also dependent on the frequency of
the sound [2].
To investigate more on this challenging experimental observation, we
developed a realistic periphery model to mimic cochlear inputs from
auditory nerves fibers onto the MSO. Using known plasticity rules such as
Spike Timing Dependent Plasticity to structure the wiring from those
connections onto the MSO, we extend the work that has already been
performed [3], and study the emergence of binaural tuning in the MSO in
a realistic scenario. The system is trained with binaural sounds such as
white noise, and then, as in most experimental papers, we tested the ITD
selectivity of cells in the MSO by presenting pure tones at various
frequencies.
Finally, we discuss, from a coding point of view the potential implications
raised by the frequency dependence of the best delay. As pointed out by
recent work [4], with such a frequency-dependent best delay, neurons in
the MSO should be seen as coding for a particular position in space,
rather than for just a fixed delay difference.
Acknowledgment: This work is funded with a Fellowship Grant from the
FRM (Fondation pour la Recherche Mdicale), and the European Research
Council (StG 240132).
References
1. Jeffress LA: A place theory of sound localization. J Comp Physiol Psychol
1948, 41(1):35-39.
2. Joris P, Yin TC: A matter of time: internal delays in binaural processing.
Trends Neurosci 2007, 30(2):70-78.
3. Fontaine B, Brette R: Neural Development of Binaural Tuning through
Hebbian Learning Predicts Frequency-Dependent Best Delays. J Neurosci
2011, 31(32):11692-11696.
4. Benichoux V, Fontaine B, Karino S, Joris P, Brette R: Frequency-dependent
time differences between the ears are matched in neural tuning (in revision).
P21
Proteomics investigation identifies prominent changes in synapse-
related proteins in a fragile X mouse model
Jantine AC Broek1*, Z Lin3, H Vant Spijker1, S Ozcan1, HM De Gruiter2,
ED Haasdijk3, R Willemsen4, CI De Zeeuw3,5, S Bahn1
1
Dept. of Chemical Engineering and Biotechnology, University of Cambridge,
Cambridge, UK; 2Erasmus Optical Imaging Center, Erasmus MC, Rotterdam,
the Netherlands; 3Dept. of Neurosciences, Erasmus MC, Rotterdam, the
Netherlands; 4Dept. of Clinical Genetics, Erasmus MC, Rotterdam, the
Netherlands; 5Netherlands Institute for Neurosciences, Royal Academy for
Arts and Sciences, Amsterdam, the Netherlands
E-mail: acb86@cam.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P21
Fragile X syndrome (FXS) is a single gene disorder that is the most
common heritable cause of intellectual disability and the most frequent
monogenic cause of autism spectrum disorders (ASD). FXS is caused by
trinucleotide repeats in the promoter region of the fragile X mental
retardation gene (fmr1). This leads to the downregulation of the fragile X
mental retardation protein (FMRP), which regulates translation of a wide
range of proteins. The extent of expression level alterations of synaptic
proteins affected by FMRP loss and their consequences on synaptic
dynamics in FXS has not been fully investigated. Here, we have
investigated the molecular mechanisms underlying FXS by identifying the
molecular signature and elucidating the function of FXS-associated
proteins. The first part of the study consists of an exploratory molecular
profiling study on brain tissue samples of a fmr1 KO mouse model using
shotgun label-free liquid chromatography mass spectrometry (LC-MSE ).
These studies indicate that most of the changed proteins are involved in
synaptic functions. Thus, as a next step we investigated synaptosomes
from the cerebellum and hippocampus using LC-MS E and label-based
selected reaction monitoring (SRM). Key findings relate to altered levels of
proteins involved in GABA-signalling in the cerebellum. To explore these
findings further, FM1-43 dye was used in cultured hippocampal neurons
and cerebellar Purkinje cells to track vesicle recycling and unloading
profiles. Furthermore, ultrastructural analysis of synaptic vesicles of the
GABA-ergic Purkinje cells in the cerebellum was performed to investigate
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
potential pre-synaptic effects. Taken together, these studies provide novel
insights into the molecular changes associated with FXS.
P22
Numerical simulations in two-dimensional neural fields
Pedro M Lima1,2*, Evelyn Buckwar1
1
Institute of Stochastics, Johannes Kepler University, 4040, Linz, Austria;
2
CEMAT/Instituto Superior Tcnico, University of Lisbon, 1049, Lisboa,
Portugal
E-mail: plima@math.ist.utl.pt
BMC Neuroscience 2015, 16(Suppl 1):P22
We present and discuss a new numerical algorithm for solving the two-
dimensional Neural Field Equation with space-dependent delays:
 overwhelming majority of neural field models use current-based synapses [1].
c V(x, t) = I(x, t) V(x, t) + K(|x y |)S(V(y, t (x, y ))dy
t 
t [0, T], x  R2
used in Neuroscience to describe the evolution of a population of
neurons and the interactions between them. A similar equation (without
delays) was first introduced by Wilson and Cowan [1], and then by Amari
[2]. Here V(x,t) represents the post-synaptic membrane potential at
instant t and position x. The function I represents external sources of
excitation and S describes the dependence between the firing rate of the
neurons and their membrane potential (typically it is a function of
sigmoidal type). The kernel function K(|x-y|) describes the connectivity
between the neurons at positions and y. The delay (x,y) takes into the
consideration the time spent by an electrical signal to travel between
these two positions.
The new numerical method presented in this work is based on an implicit
second order method for discretisation in time and uses Gaussian
quadratures for space integration. We use low-rank methods to reduce
the computational effort, which enables to reduce very significantly the
dimensions of the involved matrices, without affecting the final accuracy
of the method.
This algorithm is targeted directly to the application in Neuroscience and
Robotics. As an illustration of such applications, we consider a neural field
described in [3], where the firing rate function has the form S(x)= 2/
(1+exp(- x)) ( is a certain positive constant), and the connectivity
function is given by K(x)= 1/(2 12)1/2 exp(- ||x||2/ 2 12) - A/(2 22)1/2
exp(- ||x||2/ 2 22), with A, 1, 2 given positive constants.
It is known [3] that the stability of the trivial solution of this equation
depends on the value of and on the parameters of the connectivity
function. In particular, for each set of values of these parameters, it is
possible to compute a bifurcation value bif, such that if < bif the zero
solution is stable, and otherwise it is unstable.
With the help of our algorithm, these bifurcation values can be efficiently
computed. Some examples are displayed in Table 1 . The presented values
are for the case (x,y)=0.
Ackknowledgement: This research was supported by a Marie Curie Intra
European Fellowship, PIEF-GA-2013-629496
References
1. Wilson HR, Cowan JD: Excitatory and inhibitory interactions in localized
populations of model neurons. Biophys J 1972, 12(1):1-24.
2. Amari SL: Dynamics of pattern formation in lateral-inhibition type neural
fields. Biol Cybernet 1977, 27(2):77-87.
3. Faye G, Faugeras O: Some theoretical and numerical results for delayed
neural field equations. Physica D 2010, 239(9):561-578.
Table 1(abstract P22)
A 1 bif A 1
0 0.1 15.81 1 0.4
0 0.2 7.98 1 0.5
0 0.3 5.35 1 0.6
0 0.4 4.06 1 0.7
0 0.5 3.56 1 0.8
Page 28 of 200
P23
The neurodynamics of epilepsy: a homotopy analysis between current-
based and conductance-based synapses in a neural field model of
epilepsy
Andre DH Peterson1,2,3*, Iven MY Mareels1, Hamish Meffin4,
David B Grayden1,5, Mark J Cook2,3, Anthony N Burkitt1,5
1
NeuroEngineering Lab, Dept. of Electrical & Electronic Engineering,
University of Melbourne, Melbourne, Australia; 2Department of Medicine,
University of Melbourne, Melbourne, Australia; 3Centre for Clinical
Neurosciences, St. Vincents Hospital, Melbourne, Australia; 4NVRI, Melbourne,
Australia; 5Bionics Institute, Melbourne, Australia
E-mail: peterson@unimelb.edu.au
BMC Neuroscience 2015, 16(Suppl 1):P23
Introduction/Background: Unlike Hodgkin-Huxley type spiking models, the
Although there exist neural field models that employ conductance-based
synapses, it is not clear what their exact effects on the dynamics are,
particularly with respect to epileptic dynamics. Neural field models of epilepsy
typically describe the transition to seizure-like activity as a bifurcation [2].
This research examines the effects of conductance-based synapses on
the transition from normal to seizure-like activity in neural field models.
Methods: In this research we construct a neural field model with a
homotopy parameter, , such that when  = 0, then the model has
current-based synapses, and when  = 1, it has conductance-based
synapses. This enables us to compare and explain the key differences in
dynamics caused by the different synaptic mechanisms. In particular we
perform a bifurcation analysis using the homotopy parameter, , as a
bifurcation parameter to rigorously analyse both models.
Results: We find that the conditions under which bifurcations take place are
quite different for each synaptic mechanism. The feedback term from the
membrane potential, which makes conductance-based synapses nonlinear,
considerably affects the dynamics. This is especially so in comparison with
their linear current-based counterpart. In the current-based model
increasing the external input parameter in conjunction with the network
balance parameter generates a Hopf bifurcation, which is typically
interpreted as a transition to a seizure-like state [2]. In comparison, changing
these two parameters in the conductance-based model has no effect due to
the feedback from the membrane potential term. The suppression of the
transition to a seizure-like state occurs at a critical value of the homotopy
parameter c.
Discussion/Conclusion: In terms of neural modeling of epilepsy, this
demonstrates that the new features introduced in the more biophysically
realistic conductance-based synapses act as an anti-epileptic regulatory
mechanism that suppresses the transition to seizure. The homotopy
parameter, , is interpreted as being proportional to the synaptic
background activity that affects the conductance state of neurons and results
in fluctuations of their membrane potentials [3]. It is these fluctuations that
suppress the transition to seizure-like activity and need to be taken into
account in neural models. If these fluctuations are of reasonably small
amplitude, for example, as in resting state behaviour, then current-based
synapses can be an adequate approximation. However, if these fluctuations
are larger in amplitude, for example, as in oscillatory or seizure-like activity,
then they need to be included [4]. Essentially, we have constructed a
mathematical mapping between two different synaptic mechanisms and
examined their effects on the dynamics of a typical neural field model. This
calls into question previous results of neural field models that use current-
based synapses, including those of epilepsy, as using a more biophysically
realistic synaptic mechanism yields significantly different results.
Acknowledgements: This work was supported by ARC Linkage Project
#LP0560684 and an SVHM REF grant.
References
1. Stefanescu RA, Shivakeshavan R, Talathi SS: Computational models of
2 bif epilepsy. Seizure 2012, 21(10):748.
0.3 17.68 2. Milton J, Jung P: Epilepsy as a Dynamic Disease. Springer 2003.
3. Destexhe A, Rudolph M, Fellous JM, Sejnowski TJ: Fluctuating synaptic
0.3 10.24
conductances recreate in vivo-like activity in neocortical neurons.
0.3 8.41 Neuroscience 2001, 107(1):13-24.
0.3 7.94 4. Richardson M: Firing-rate response of linear and nonlinear integrate-and-
fire neurons to modulated current-based and conductance-based
0.3 8.09
synaptic drive. Phys Rev E 2007, 76:021919.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
P24
Incremental stability of delayed neural fields: a unifying framework for
endogenous and exogenous sources of pathological oscillations
Georgios Is. Detorakis1,2*, Antoine Chaillet1,2
1
University Paris Sud, Orsay, Paris, 91400, France; 2LSS, Suplec, Gif sur Yvette,
Paris, 91190, France
E-mail: georgios.detorakis@lss.supelec.fr
BMC Neuroscience 2015, 16(Suppl 1):P24
Neural fields are integro-differential equations that have been extensively
used to model spatiotemporal evolution of neocortical areas (see [1] for a
detailed review). Time-delayed neural fields have also been a matter of
investigation since they take into account axonal delays [2]. On the other
hand, time-delay finite dimensional systems have been used in models of
Parkinsons disease: delays have been shown to play a possible role in the
generation of pathological neural oscillations linked to motor symptoms of
Parkinson disease in a firing-rate model of basal ganglia [3,4]. Nonetheless,
these models fail at rendering the spatial distribution of the neural activity
of the populations involved. Two possible mechanisms for the onset of
pathological oscillations in basal ganglia have been investigated in the
literature. The first one, the endogenous mechanism, hypothesizes that
dopamine depletion tends to increase the synaptic gains between the
excitatory neurons of the subthalamic nucleus (STN) and the inhibitory
neurons of the external segment of globus pallidus (GPe), thus generating
an instability that translates into sustained oscillations. The second one,
the exogenous mechanism, explains these oscillations onset by a
diffusion of spontaneous oscillations from external structures (such as
Striatum) to the GPe-STN network [5].
The main goal of this work is to deepen this analysis by providing
theoretical conditions under which a network of time-delayed neural field
equations is incrementally stable. We believe that incremental stability
constitutes an instrumental framework to investigate both the
mechanisms evoked above. Indeed, by considering constant inputs to the
basal ganglia, incremental stability ensures convergence to a unique
equilibrium configuration, thus ruling out the possibility of endogenous
mechanism for oscillations onset. On the other hand, incremental stability
guarantees entrainability to periodic inputs (meaning convergence to a
T-periodic solution in response to any T-periodic input), and can thus be
useful to unravel the mechanism of pathological diffusion from external
structures in the exogenous scenario.
Relying on the Razumikhin-Lyapunov approach here we derive these
sufficient conditions for incremental stability of delayed neural fields. This
theoretical framework thus complements the Krasovskii-Lyapunov
approach already used in the literature to address the stability of delayed
neural fields equations [6]. Simulations confirm our theoretical
expectations and demonstrate that interconnected neural fields can
exhibit sustained oscillations, according to either the endogenous or the
exogenous mechanism, depending on the strength of the synaptic
weights between the excitatory (STN) and the inhibitory (GPe)
populations. The derived theoretical results thus seem to constitute a
fertile ground for further investigations based on experimental data, to
discriminate between the endogenous and the exogenous hypotheses
for Parkinsonian sustained oscillations in the STN-GPe network.
Acknowledgements: This work has received support from ANR JCJC
SynchNeuro and from the iCODE institute project funded by the IDEX
Paris-Saclay, ANR-11-IDEX-0003-02.
References
1. Bressloff PC: Spatiotemporal dynamics of continuum neural fields. Journal
of Physics A: Mathematical and Theoretical 2011, 45(3):033001.
2. Veltz R, Faugeras O: Stability of the stationary solutions of neural field
equations with propagation delays. The Journal of Mathematical
Neuroscience 2011, 1(1):1-28.
3. Nevado Holgado AJ, Terry JR, Bogacz R: Conditions for the Generation of
Beta Oscillations in the Subthalamic Nucleus-Globus Pallidus Network.
Journal of Neuroscience 2010, 30(37):12340-12352.
4. Haidar I, Pasillas-Lepine W, Panteley E, Chaillet A, Palfi S, Senova S: Analysis
of delay-induced basal ganglia oscillations: the role of external
excitatory nuclei. International Journal of Control 2014, 80(8):1936-1956.
5. McCarthy M, Moore-Kochlacs C, Gu X, Boyden ES, Han X, Kopell N: Striatal
origin of the pathologic beta oscillations in Parkinsons disease. PNAS
2011, 108(28):11620-11625.
Page 29 of 200
6. Faye G, Faugeras O: Some theoretical and numerical results for delayed
neural field equations. Physica D 2010, 239(9):561-578.
P25
A spiking network model of basal ganglia to study the effect of
dopamine medication and STN-DBS during probabilistic learning task
Alekhya Mandali*, V Srinivasa Chakravarthy
Department of Biotechnology, Bhupat and Mehta School of Biosciences,
Chennai, 600036, Tamil Nadu, India
E-mail: schakraiitm@ac.in
BMC Neuroscience 2015, 16(Suppl 1):P25
The effect of dopamine medication (L-Dopa and Dopamine (DA) agonists)
and Deep Brain Stimulation (DBS) of Sub Thalamic Nucleus (STN) on the
cognition of Parkinsons disease (PD) patients has been of interest to
neuroscientists, owing to their ability to produce impulsive behavior as a
side effect. Interestingly, it has been found that the impulsive decision
making in STN-DBS patients has been observed to be quite contrary. As an
attempt to understand the aforementioned side effect, we built a spiking
network model of basal ganglia (BG) and tested on a probabilistic learning
task [1]. BG nuclei such as Globus Pallidus externa (GPe), interna (GPi) and
STN were represented as 2D arrays of Izhikevich neurons (50 50 lattice)
and the activity of Striatal neurons (D1 and D2 receptor expressing
Medium Spiny Neurons) as a Poisson process. Invoking the idea that DA
cells in Substantia Nigra par compacta code for reward prediction error,
similar to the temporal difference (TD) error term in Reinforcement
learning, learning was introduced by updating the cortico-striatal weights
using TD error. In the probabilistic learning task, the system was subjected
to 3 pairs of stimuli (one at a time) randomly and the goal is to learn to
select the most rewarding stimulus (=A) and avoid the punishing one (=B).
The task had both training and testing stages where the model had to
learn to make an optimal decision. The models performance was tested
on Healthy controls, PD in OFF, PD ON (L-Dopa & DA agonists) and finally
on STN-DBS in terms of percentage accuracy and Reaction time (RT).
Results: The performance PD-OFF cases showed skewness towards
punishment learning (avoiding B), whereas the opposite was observed in
PD-ON (L-Dopa) case (fig. 1A). These results are consistent with
experimental results which suggest that the low (high) DA in PD-OFF
(ON) state leads to higher punishment (reward) sensitivity. Our model
results are also consistent with other experimental studies related to DA
agonist effects on decision making. The lowest choice reaction times for
PD-DA agonist condition (Fig 1B) (among all other cases) indicates
impulsivity leading to suboptimal action selection observed in decreased
% Accuracy levels (Fig. 1A). The effect of DBS electrode position on RT
and % accuracy levels (Fig. 1C & 1D) were found also to be interesting,
where a change in electrode position made the model performance to
switch from punishment learning to reward learning. These results
suggest that the DBS electrode position might play a major role in
inducing impulsivity in the PD patients.
Reference
1. Frank MJ, Samanta J, Moustafa AA, Sherman SJ: Hold your horses:
impulsivity, deep brain stimulation, and medication in Parkinsonism.
Science 2007, 318(5854):1309-1312.
P26
Modulation of neural firing through intracellular ATP dynamics
governed by energy feedback from the vascular system
Karishma Chhabria, V Srinivasa Chakravarthy*
Department of Biotechnology, Indian Institute of Technology Madras,
Chennai, 600036, Tamil Nadu, India
E-mail: schakra@ee.iitm.ac.in
BMC Neuroscience 2015, 16(Suppl 1):P26
We propose a simple model for neuro-glio-vascular interactions to
emphasize on the role of energy feedback from the vascular system in
brains computations [1,2]. In [1], we introduced a bidirectional
communication within a detailed biophysical model of neuron-astrocyte-
vessel. We now compress this model to just two modules: the neuron and
the energy module. The energy module is a lumped representation of the
astrocyte-vessel system; it receives neural firing activity as input and controls
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 30 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

Figure 1(abstract P25) Percentage Accuracy and Reaction times for various simulated of network model. (A) % percentage accuracy (B) Reaction
Times for Low and High conflict conditions. (C) % Accuracy levels for different positions of DBS electrode in STN (D) and the corresponding Reaction
times

Figure 1(abstract P26) A. Implemented network configuration with n (=1000) neurons, i synapses with c (=20%) connectivity. B. Simulated LFP at
both high (=10) and low (=3) p. C. & D. depicts spectrograms for low and high p, respectively

intracellular neuronal energy (ATPi) levels as a feedback. The model
comprises of a quadratic integrate and fire neuron with a dynamic
threshold, Vth, which further depends on the ATPi dynamics. Vth is high
during ATPi deficit, making the neuron least excitable and vice versa for
high ATPi conditions. The underlying principle of modeling Vth as a function
of ATPi is based on the experiments describing the role of KATP channels in
governing neural excitability [3]. These channels are ATP-dependent
potassium channels and are open when ATPi is low, resulting in a
depolarized membrane potential of the neuron during metabolically
compromised states such as hypoxia [3].
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
The neuron model parameters are adapted to that of mammalian cortical
pyramidal neuron [4]. Furthermore, ATPi dynamics are also modeled
similar to [4], where ATPi consumption directly depends on neural spiking
activity. The production rate of ATPi, p is a crucial model parameter
representing the local vascular activity. A wide range of neural dynamic
behaviors: phasic bursting, tonic bursting and continuous spiking are
observed by varying p and external input current Iext. Furthermore,
simulation of a network consisting of such energy-dependent neural units
(Figure 1A.) depicts that p could modulate the Local field potential (LFP)
frequencies and amplitudes (Figure 1B.). Interestingly, low frequency LFP
dominates under low p conditions and could represent seizure-
like activity observed in epilepsy. Although conventional neuroscience
considers unidirectional influences from neurons to small vessels, there
have been proposals that highlight the reverse influence from the vessels
to the neurons [1,2] . The proposed neuron-energy unit may be treated
as a building-block in large scale models of neurovascular networks.
References
1. Chander BS, Chakravarthy VS: A computational model of neuro-glio-
vascular loop interactions. PloS One 2012, 7(11):e48802.
2. Moore CI, Cao R: The hemo-neural hypothesis: on the role of blood flow
in information processing. Journal of Neurophysiology 2008, 99(5):2035.
3. Wang L, Zhu Q-L, Wang G-Z, Deng T-Z, Chen R, Liu M-H, Wang S-W: The
protective roles of mitochondrial ATP-sensitive potassium channels
during hypoxia-ischemia-reperfusion in brain. Neuroscience Letters 2011,
491(1):63-67.
4. Ching S, Purdon PL, Vijayan S, Kopell NJ, Brown EN: A neurophysiological-
metabolic model for burst suppression. Proceedings of the National
Academy of Sciences 2012, 109(8):3095-3100.
P27
Could the prior development of the retinotopic map account for the
radial bias in the orientation map in V1?
Ryan Thomas Philips*, V Srinivasa Chakravarthy
Department of Biotechnology, Indian Institute of Technology Madras,
Chennai, 600036, Tamil Nadu, India
E-mail: schakra@iitm.ac.in
BMC Neuroscience 2015, 16(Suppl 1):P27
The development of the retinotopic map is presumed to precede the
development of the orientation map in V1 in primates. Experimental studies
demonstrate a radial bias, wherein radial orientations produce higher
activity compared to other orientations [1]. However most traditional models
assume isometry in the orientation map developed, i.e. the orientation maps
throughout V1 are similar in nature, independent of its retinotopy. In
this paper, we propose an activity-dependent model which simulates
the development of a radially biased orientation map. To that end we
simulate the large-scale development of the retinotopic map, followed by
the development of the orientation map in a sub-region of this map. The
architecture consists of a Laterally Interconnected Synergetically Self
Organizing Map (LISSOM) [2] with 2 layers, representing the retina, and the
V1 respectively (see Figure 1A). At each time step, each neuron in V1,
Figure 1(abstract P27) (A) Schematic representation of the LISSOM architecture. (B) V1 orientation map developed after initial training to establish
retinotopy along with its color map. (C) Orientation sub-maps, biased by initial retinotopy, at 1000, 2000, 3000, 4000 iterations. (D) Orientation sub-maps
at 1000, 2000, 3000, 4000 iterations assuming isometry. (E) Histogram of the area covered by each of the orientations (color coded) corresponding to (C).
(F) Histogram of the area covered by each of the orientations (color coded) corresponding to (D)
Page 31 of 200
combines the afferent activation (r1,r2) along with its lateral excitations and
inhibitions (hkl) from the previous time step.
  
ij (t) = ( r1,r2 ij,r1r2 + E Eij,kl kl (t 1) I Iij,kl kl (t 1)) (1)
r1,r2 k,l k,l
The afferent (ij,r1r2), lateral excitatory (Eij,kl) and lateral inhibitory (Iij,kl)
weights adapt based on a normalized Hebbian mechanism. In order to
develop the retinotopic map, the inputs to the retinal layer consists of
centered (assumed to be the point of fixation) rectangular bars of varying
dilations and rotations as modelled in [3]. The retinotopic map
developed, biases the initial configuration of the orientation map (see
Figure 1B) since all the bars given during the initial training are centered.
For the subsequent refinement of the orientation map, Gaussians of
differing orientation and positions (non-centered) are given as inputs to
the retinal layer. After training for 4000 iterations (see Figure 1C,D), it is
observed that the developed orientation map prefers those orientations
which the retinotopy biases it towards, quantified by their corresponding
histograms (See Figure 1E,F). As seen from the histogram the area
occupied by the region mapping 1250-1500 is larger in the map
developed assuming retinotopic bias, compared to that of the map
developed assuming isotropy.
Conclusions: A neural activity based model for the development of
radially biased orientation maps in V1 is demonstrated.
References
1. Sasaki Y, Rajimehr R, Kim BW, Ekstrom LB, Vanduffel W, Tootell RB: The
radial bias: a different slant on visual orientation sensitivity in human
and nonhuman primates. Neuron 2006, 51(5):661-670.
2. Bednar JA: Topographica: building and analyzing map-level simulations
from Python, C/C++, MATLAB, NEST, or NEURON components. Frontiers
in Neuroinformatics 2009, 3:8.
3. Philips R, Chakravarthy S: The mapping of eccentricity and meridional
angle onto orthogonal axes in the primary visual cortex: An activity-
dependent developmental model. Frontiers in Computational Neuroscience
2015, 9:3.
P28
Could the prior development of the retinotopic map account for the
radial bias in the orientation map in V1?
Ryan Thomas Philips*, V Srinivasa Chakravarthy
Department of Biotechnology, Indian Institute of Technology Madras,
Chennai, 600036, Tamil Nadu, India
E-mail: schakra@iitm.ac.in
BMC Neuroscience 2015, 16(Suppl 1):P28
The development of the retinotopic map is presumed to precede the
development of the orientation map in V1 in primates. Experimental studies
demonstrate a radial bias, wherein radial orientations produce higher
activity compared to other orientations [1]. However most traditional models
assume isometry in the orientation map developed, i.e. the orientation maps
throughout V1 are similar in nature, independent of its retinotopy. In this
paper, we propose an activity-dependent model which simulates the
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 32 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

Figure 1(abstract P28) (A) Schematic representation of the LISSOM architecture. (B) V1 orientation map developed after initial training to establish
retinotopy along with its color map. (C) Orientation sub-maps, biased by initial retinotopy, at 1000, 2000, 3000, 4000 iterations. (D) Orientation sub-maps
at 1000, 2000, 3000, 4000 iterations assuming isometry. (E) Histogram of the area covered by each of the orientations (color coded) corresponding to (C).
(F) Histogram of the area covered by each of the orientations (color coded) corresponding to (D)

development of a radially biased orientation map. To that end we simulate
the large-scale development of the retinotopic map, followed by the
development of the orientation map in a sub-region of this map. The
architecture consists of a Laterally Interconnected Synergetically Self
Organizing Map (LISSOM) [2] with 2 layers, representing the retina, and the
V1 respectively (see Figure 1A). At each time step, each neuron in V1,
combines the afferent activation (r1,r2) along with its lateral excitations and
inhibitions (hkl) from the previous time step.
 
ij (t) = ( r1,r2 ij,r1r2 + E Eij,kl kl (t 1) I Iij,kl kl (t 1))
r1,r2 k,l
The afferent (ij,r1r2), lateral excitatory (Eij,kl) and lateral inhibitory (Iij,kl)
weights adapt based on a normalized Hebbian mechanism. In order to
develop the retinotopic map, the inputs to the retinal layer consists of
centered (assumed to be the point of fixation) rectangular bars of varying
dilations and rotations as modelled in [3]. The retinotopic map developed,
biases the initial configuration of the orientation map (see Figure 1B) since all
the bars given during the initial training are centered. For the subsequent
refinement of the orientation map, Gaussians of differing orientation and
positions (non-centered) are given as inputs to the retinal layer. After training
for 4000 iterations (see Figure 1C,D), it is observed that the developed
orientation map prefers those orientations which the retinotopy biases
it towards, quantified by their corresponding histograms (See Figure 1E,F).
As seen from the histogram the area occupied by the region mapping 1250-
1500 is larger in the map developed assuming retinotopic bias, compared to
that of the map developed assuming isotropy.
Conclusions: A neural activity based model for the development of
radially biased orientation maps in V1 is demonstrated.
References
1. Sasaki Y, Rajimehr R, Kim BW, Ekstrom LB, Vanduffel W, Tootell RB: The
radial bias: a different slant on visual orientation sensitivity in human
and nonhuman primates. Neuron 2006, 51(5):661-670.
2. Bednar JA: Topographica: building and analyzing map-level simulations
from Python, C/C++, MATLAB, NEST, or NEURON components. Frontiers
in Neuroinformatics 2009, 3:8.
3. Philips R, Chakravarthy S: The mapping of eccentricity and meridional angle
onto orthogonal axes in the primary visual cortex: An activity-dependent
developmental model. Frontiers in Computational Neuroscience 2015, 9:3.
 P29
k,l
(1) A model of learning temporal delays, representative of adaptive
myelination
Meenakshi Asokan1, Karishma Chhabria2, V Srinivasa Chakravarthy2*
1
Department of Electrical Engineering, Indian Institute of Technology Madras,
Chennai, 600036, Tamil Nadu, India; 2Department of Biotechnology, Indian
Institute of Technology Madras, Chennai, 600036, Tamil Nadu, India
E-mail: schakraee@iitm.ac.in
BMC Neuroscience 2015, 16(Suppl 1):P29
Learning and plasticity in the brain has been generally attributed to the
synaptic activity in a neuronal network. However, recent studies [1] propose
that the changes in conduction velocity of action potentials could affect
the synchrony of spike arrival timings at the synapse, thereby modulating
plasticity. This is attributed to adaptive myelination brought about by the
oligodendrocytes (a class of glia that myelinate the axons in the central
nervous system). We propose that the temporal delays in a neuronal
network could be trained in addition to the training solely synaptic weights,
in response to dynamic input spike patterns. These temporal delays are
trained using the Spike Timing Dependent Plasticity (STDP) kernel, which is
a temporally asymmetric variant of Hebbian learning. This paradigm of
modeling is motivated from a study which describes that in addition to the
pre-synaptic activity, oligodendrocytes can sense the post synaptic activity
relayed through the astrocyte activity [2].

Figure 1(abstract P29) A. Network architecture B. Pictorial representation of the input patterns (first layer) C. Output of the SOM (second layer)
corresponding to each input pattern D. Processed second layer output is fed to the third layer to train the delays () and the weights (w)
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
The proposed model comprises of three layers (Figure 1A.); the first layer
represents the input (different bar orientations and corresponding spatial
locations (Figure 1B.)) to the Self Organizing Map (SOM) (second layer). For
every bar orientation, a different neuron in the SOM is activated for each
spatial location (Figure 1C.). This sequence of static outputs are cascaded
depending on the direction of motion for each orientation and fed to the
third layer as dynamic spike trains (Figure 1D.). The weights between the
second and the third layer are trained by Hebbian learning and normalized
after each input presentation. Furthermore, the delays are simultaneously
trained using the STDP algorithm wherein the pre-synaptic spikes are input
spike trains, time shifted by temporal delays. The post synaptic spikes are
calculated by integrating the Post Synaptic Potentials (PSPs), for a given
threshold and the neuron having the maximum amplitude of the
integrated PSP is chosen as the winner. Simulation of such a network
results in different neurons activated in response to motions of bars of
different orientations. In contemporary neural network studies, temporal
delays are typically ignored or held constant. However, the plasticity of
conduction delays adds a novel dimension to the study of neural
information processing. Moreover, future exploration in this domain could
possibly explain the correlations of hyper and hypo synchrony of neural
firing with disorders such as dyslexia and schizophrenia [3].
References
1. Fields RD: Oligodendrocytes changing the rules: action potentials in glia
and oligodendrocytes controlling action potentials. Neuroscientist 2008,
14(6):540-543.
2. Ishibashi T, Dakin KA, Stevens B, Lee PR, Kozlov SV, Stewart CL, Fields RD:
Astrocytes promote myelination in response to electrical impulses.
Neuron 2006, 49(6):823-832.
3. Pajevic S, Basser PJ, Fields RD: Role of myelin plasticity in oscillations and
synchrony of neuronal activity. Neuroscience 2014, 276:135-147.
P30
Identifying excitatory and inhibitory synapses in neuronal networks
from dynamics using Transfer Entropy
Felix Goetze1,2*, Pik-Yin Lai1, CK Chan1,3
1 mechanism of MN neuroprotection [1] and therefore we hypothesize
Department of Physics, National Central University, Chung-Li, Taiwan,
Republic of China; 2Taiwan International Graduate Program for Molecular
Science and Technology, Institute for Atomic and Molecular Sciences,
Academia Sinica, Taipei, Taiwan, Republic of China; 3Institute of Physics,
Academia Sinica, Taipei, Taiwan, Republic of China
E-mail: afgoetze@gmail.com
BMC Neuroscience 2015, 16(Suppl 1):P30
Measuring the effective connectivity between the elements of a neuronal
system promises to give conclusions about brain function and the
principles of information processing in the brain. Transfer Entropy [1] (TE)
and its extensions have recently become more and more popular for this
task and have been applied to various types of neuronal data, such as
from EEG, Calcium Imaging and Multi-electrode Array measurements [2].
Being based on information theory, TE can be interpreted as the predictive
information transfer between two time series. TE is a model free
measurement and quantifies even non-linear interactions and very
importantly the directionality.
However, due to its information-theoretic nature, TE doesnt distinguish
between different types of interactions, for example whether a pre-
synaptic neuron drives a post-synaptic neuron via an excitatory or an
inhibitory synapse [3]. This distinction is crucial for the understanding of
the network dynamics and the exact interplay of excitation and inhibition
in neuronal networks plays an important role for network bursts and
synchronization. The balance of excitation and inhibition is believed to
play a role for the occurrence of epileptic seizures [4].
We propose a method complementary to the TE measurement, to not only
measure the effective connectivity from neuronal network dynamics, but also
to classify, whether the interactions are excitatory or inhibitory. We achieve
this by introducing a new quantity, which is a linear combination of the
individual terms that sum up to the Transfer Entropy. This quantity can be
computed by applying a Principal Component Analysis (PCA) to the Transfer
Entropy terms across different measurements. It assumes positive values for
excitatory connections and negative values for inhibitory connections.
To verify this method we apply it to the spike trains of noise-driven recurrent
neuronal network motifs, simulated with an extended Fitzhugh-Nagumo
Page 33 of 200
model. We analyze all possible combinations of three neurons being
coupled with either excitatory or inhibitory synapses, which add up to a
total of 132 network motifs. The best performance of our reconstruction of
these motifs, being measured by the area under the Receiver Operating
Characteristic curve, exceeds values of 0.99. Together with the PCA analysis,
we are not only able to reconstruct the motif topologies from the neuronal
dynamics, but we are also able to distinguish between excitatory and
inhibitory synapses.
References
1. Schreiber T: Measuring information transfer. Physical Review Letters 2000,
85:461.
2. Wibral M, Vicente R, Lizier JT: Directed Information Measures in Neuroscience
2014.
3. Orlandi JG, Stetter O, Soriano J, Geisel T, Battaglia D, Garcia-Ojalvo J:
Transfer Entropy Reconstruction and Labeling of Neuronal Connections
from Simulated Calcium Imaging. PLoS One 2014, 9(6):e98842.
4. McCormick DA, Contreras D: On the Cellular and Network Bases of
Epileptic Seizures. Annual Review of Physiology 2001, 63:815-846.
P31
Early dysregulation of trigeminal motor pool excitability in a mouse
model for neurodegenerative motoneuron disease
Sharmila Venugopal1*, Martina Wiedau-Pazos2, Scott H Chandler1
1
Department of Integrative Biology and Physiology, David Geffen School of
Medicine, University of California Los Angeles, Los Angeles, CA, USA;
2
Department of Neurology, David Geffen School of Medicine, University of
California Los Angeles, Los Angeles, CA, USA
E-mail: vsharmila@ucla.edu
BMC Neuroscience 2015, 16(Suppl 1):P31
Amyotrophic Lateral Sclerosis (ALS) is a progressive neurodegenerative
motoneuron (MN) where in fast fatigable motor units (MUs) of
vulnerable motor pools preferentially degenerate followed by fast
fatigue resistant and slow MUs . Excitability is a key endogenous
that pre-symptomatic excitability indicates impending disease
development. Using a transgenic mouse model for ALS, we performed
in vitro patch-clamp electrophysiology in ALS-vulnerable trigeminal
motoneurons (TMNs) retrogradely labeled from jaw closer muscles at
P8-12. We proposed a novel k-means clustering approach to classify
TMNs into putative fast fatigable (PFF), fast fatigue resistant (PFR) and
slow (PS) MUs based on rheobase and input resistance. Interestingly,
hyper-excitability was noted in PFF and PFR TMNs (Fig. 1B) while hypo-
excitability was evident in a subset of PS TMNs with linear frequency-
current (F-I) characteristics compared to wild-type TMNs [2]. The F-I
relationships displayed dysregulation across the motor pool (Fig. 1C).
A jaw closer motor pool model was developed and simulated in
MATLAB using observed alterations in membrane properties amongst
the mutant MU types. Model results predict resistance in muscle force
initiation and reduced motor pool dynamic range due to opposite
changes in excitability between slow and fast TMNs (Fig. 1D).
Acknowledgements: This study was supported by NIH grant NS071348
to S.H.C
References
1. Saxena S, Roselli F, Singh K, Leptien K, Julien JP, Gros-Louis F, et al:
Neuroprotection through excitability and mTOR required in ALS
motoneurons to delay disease and extend survival. Neuron 2013,
80(1):80-96.
2. Venugopal S, Hsiao C, Sonoda T, Weida-Pazos M, Chandler S: Homeostatic
dysregulation in membrane properties of masticatory motoneurons
compared to oculomotor neurons in a mouse model for Amyotrophic
Lateral Sclerosis. Journal of Neuroscience 2015, 35(2):707-720.
P32
Non-invasively recorded transient pathological high-frequency
oscillations in the epileptic brain: a novel signature of seizure evolution
Catherine Stamoulis1,2*, Bernard Chang1,3
1
Harvard Medical School, Boston, MA, 02115, USA; 2Departments of
Radiology and Neurology, Boston Childrens Hospital, Boston, MA, 02115,
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P31) A. Classification of TMNs into PFF, PFR and PS MU types using k-means clustering based on rheobase and input
resistance (Rinp). B. Rheobase values were significantly lower among mutant PFR and PFF units compared with wild-type (Students t test, *p = 0.0002
for PFR, **p = 0.0001 for PFF), whereas Rinp values were significantly greater (Students t test, *p = 0.0009 for PFR, **p = 0.0028 for PFF). Error bars indicate
SD. C. F-I relationships of PS, PFR and PFF TMNs in wild-type and mutant mice; color coding is similar to A. D. Simulation of isometric force versus MN
input current across the jaw closer motor pool. Dashed lines demarcate normal (gray) and altered (red) muscle dynamic range
USA; 3Department of Neurology, Beth Israel Deaconess Medical Center,
Boston, MA, 02215, USA
E-mail: caterina.stamoulis@childrens.harvard.edu
BMC Neuroscience 2015, 16(Suppl 1):P32
Background: Brain activity is characterized by oscillatory patterns that occur
at a wide range of frequencies and have been associated with various
cognitive processes and behaviors. Depending on their dominant frequency,
brain oscillations facilitate the coordinated activation of neuronal networks in
response to external inputs and cognitive demands. The vast majority of
human studies have focused on neural oscillations at frequencies typically
less than 50 Hz. In contrast, relatively few studies have investigated neural
activity above 50-60 Hz, in part due to the fact that relatively low-amplitude
oscillations at these frequencies are difficult to identify in human
electrophysiological signals, particularly from the intact brain. Both
physiological and pathological oscillations at frequencies >80 Hz have been
previously reported, but their occurrence, dynamics and role remain elusive.
Pathological high-frequency oscillations (HFO) detected in invasive
recordings from the epileptic brain may be correlated with epileptogenic
brain tissue and may provide insights into the physiology of the
hyperexcitable brain. While recent studies have shown that HFOs are also
detectable in scalp electroencephalography (EEG) [1], very little is known
about their physiological origin and relationship to seizure dynamics.
Methods: This study investigated the dynamics of scalp-recorded transient
high-frequency oscillations (sHFO) identified in continuous multi-day EEG
recordings from 12 epilepsy patients with medically intractable seizures.
Using a time-domain decomposition approach for non-stationary signals,
individual EEG signals were decomposed into their dominant components,
including low-amplitude non-random, non-artifact signals with characteristic
frequencies >80 Hz. Based on their frequency and waveform durations these
corresponded to transient sHFOs.
Results: Transient, low-amplitude (typically <20 V) oscillations with durations
of <50 - ~100 ms and dominant frequencies in the range >80-190 Hz were
identified in continuous scalp EEG in both wakefulness and sleep. Based on
their frequencies these oscillations fell into two distinct ranges ~80-125 Hz
and ~150-195 Hz, and occurred intermittently over long periods of time (of
the order of several hours). Seizures consistently occurred during intervals of
increased sHFO amplitude and decreased sHFO frequency and were often
time-locked to local changes in sHFO parameters. Finally, sHFOs were found
to be spatially localized both within and beyond the epileptogenic region,
with asymmetric spatial patterns that were more localized in interictal,
immediate preictal and postictal intervals and more spatially distributed
during ictal epochs, possibly associated with seizure propagation.
Conclusions: Low-amplitude sHFOs were consistently detected in
continuous EEG often hours prior to ictal onset. These oscillations had
Page 34 of 200
short durations, their temporal patterns were independent of the sleep-
wake cycle, and were detected primarily in electrodes covering the
epileptogenic region as well as a few contralateral electrodes. The majority
of seizures occurred during intervals of increased sHFO amplitude,
suggesting that these oscillations may be associated with a relatively long
period of ictogenesis. Consequently, sHFOs may represent a novel
electrophysiological signature of the epileptic brain, and may be associated
with local generators of abnormal neural activity beyond the epileptogenic
region. These findings further suggest that sHFO may dynamics may be
used for targeted clinical intervention to prevent seizure occurrence.
Reference
1. Stamoulis C, Gruber LJ, Schomer DL, Chang BS: High-frequency neuronal
network modulations encoded in scalp EEG precede the onset of focal
seizures. Epilepsy Behav 2012, 23(4):471-480.
P33
Neuromechanical bistability contributes to robust and flexible behavior
in a model of motor pattern generation
David N Lyttle1,2*, Jeffrey P Gill1, Kendrick M Shaw1, Peter J Thomas2,
Hillel J Chiel1
1
Department of Biology, Case Western Reserve University, Cleveland, OH
44106, USA; 2Department of Mathematics, Applied Mathematics, and
Statistics, Case Western Reserve University, Cleveland, OH 44106, USA
E-mail: dnl13@case.edu
BMC Neuroscience 2015, 16(Suppl 1):P33
Motor systems must be both robust (able to generate behavior reliably
despite perturbations) and flexible (able to adapt to a variable environment).
How can motor systems be insensitive to external perturbations, while
simultaneously retaining sensitivity to sensory input when it is necessary?
One possibility is that motor systems respond to changing environmental
demands by switching between different stable dynamical regimes, each
having different sensitivities to sensory input and producing different
behaviors (multifunctionality). Multifunctional motor systems have been
studied in many species, and prior work suggests it can arise from
multistable dynamics in the nervous system and/or the kinematics of the
body [1,2]. Here we explore how multistability can enable a motor system to
rapidly adjust its sensitivity to mechanosensory perturbations, thereby
allowing it to shift the balance between robustness and flexibility as needed.
Specifically, we demonstrate the coexistence of multiple stable dynamical
regimes with different sensitivities to sensory inputs in a neuromechanical
model, and explore the functional consequences of this multistability.
We focus on a nominal neuromechanical model of the feeding apparatus
of the marine mollusk Aplysia californica as a concrete example [3].
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
In this model, the dynamics of three mutually inhibitory neural pools yield
a stable heteroclinic channel, in which a stable limit cycle passes near
some number of saddle equilibrium points, thus creating regions of
localized slowing where sensory inputs can prolong specific phases of the
cycle [4,5]. The neural dynamics drive a biomechanical model of the
Aplysia buccal mass, and proprioceptive feedback from the feeding
apparatus is in turn sent back to the nervous system. The model is
confronted with environmental variability in the form of a simulated
feeding task: On each behavioral cycle, the feeding apparatus attempts to
grasp a length of seaweed (represented as a fixed mechanical load), but
only succeeds in doing so with some probability. Once a piece of seaweed
has been successfully grasped, the system must consume the entire
seaweed strip before attempting to grasp another one. Overall
performance is assessed by averaging the net seaweed consumption over
many trials.
We find that the model exhibits bistability between two distinct regimes: a
heteroclinic regime and a limit cycle regime. In the heteroclinic regime, Acknowledgements: This work was supported by funding from the
sensory feedback is able to selectively slow the neural dynamics, thus
providing the (slower) muscles with sufficient time to adapt to mechanical
loads. In the limit cycle regime, the dynamics are insensitive to
proprioceptive feedback, and the oscillations are faster. We find that the
limit cycle regime is better at grasping seaweed due to the increased
frequency of cycling, whereas the heteroclinic regime performs better with
respect to consuming seaweed once it has been grasped, due to the
enhanced sensitivity to sensory input. Moreover, the ability of the model
to flexibly transition between regimes allows it to outperform variants of
the model that can only operate in one regime or the other. Thus,
multistable dynamics that arise as a consequence of brain-body coupling
may allow motor systems to produce robust and flexible behaviors in
variable environments.
References
1. Golubitsky M, Stewart I, Buono PL, Collins JJ: Symmetry in locomotor
central pattern generators and animal gaits. Nature 1999,
401(6754):693-695.
2. Sutton GP, Mangan EV, Neustadter DM, Beer RD, Crago PE, Chiel HJ: Neural
control exploits changing mechanical advantage and context
dependence to generate different feeding responses in Aplysia. Biol
Cybern 2004, 91(5):333-345.
3. Shaw KM, Lyttle DN, Gill JP, Cullins MJ, McManus JM, Lu H, Thomas PJ,
Chiel HJ: The significance of dynamical architecture for adaptive
responses to mechanical loads during rhythmic behavior. J Comput
Neurosci 2015, 38(1):25-51.
4. Shaw KM, Park YM, Chiel HJ, Thomas PJ: Phase Resetting in an
Asymptotically Phaseless System: On the Phase Response of Limit Cycles
Verging on a Heteroclinic Orbit. SIAM J Appl Dyn Syst 2012, 11(1):350-391.
5. Rabinovich MI, Huerta R, Varona P, Afraimovich VS: Transient Cognitive
Dynamics, Metastability, and Decision Making. PLoS Comput Biol 2008,
4(5):e1000072.
P34
Investigating the effects of beta-amyloid on hippocampal signalling in
Alzheimers disease
Julia M Warburton1*, Daniel J Whitcomb2,3, Krasimira Tsaneva-Atanasova4,
Kei Cho2,3
1
Bristol Centre for Complexity Sciences, University of Bristol, Bristol, BS8 1TR,
UK; 2Henry Wellcome Laboratories for Integrative Neuroscience and
Endocrinology (HW-LINE), University of Bristol, Bristol, BS1 3NY, UK; 3Centre
for Synaptic Plasticity, University of Bristol, Bristol, BS1 3NY, UK; 4Department
of Mathematics, University of Exeter, Exeter, EX4 4QF, UK
E-mail: julia.warburton@bristol.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P34
Alzheimers disease (AD) is the leading form of dementia and is
characterised clinically by cognitive decline and impairments to memory
function. The protein amyloid-b (Ab) is thought to be a key mediator of this
neurodegeneration [1]. Mounting evidence suggests that soluble Ab1-42
oligomers (oAb1-42) cause the impairments observed in early AD, which
includes synapse loss and synaptic dysfunction. For example, oAb1-42 inhibits
hippocampal long-term potentiation (LTP) [2], an important synaptic
plasticity thought to underlie memory formation in the brain [3]. We have
recently found that postsynaptic infusion of oAb into hippocampal neurons
Page 35 of 200
causes an increase in synaptic conductance attributed to changes in the
excitatory glutamate receptor profile at the synapse. The relevance of this to
synaptic dysregulation and consequence on hippocampal signalling is yet to
be determined.
Here we take a computational approach to investigate how oAb 1-42
dysregulates signalling within the hippocampus. We first develop a
biophysical model of synaptic potentiation in a CA1 neuron, based on a
simple kinetic synapse model and Hodgkin-Huxley formalism [4], which
successfully reproduces the conductance increase observed in single-cell
patch clamp experiments following oAb application. We then extend this to
investigate how this affects signalling across a small network of CA1
neurons, analysing the importance of connectivity strength and number of
affected neurons to the overall dynamics of the system. Recent studies have
suggested that there are two distinct types of CA1 neurons with different
electrophysiological properties [5]. Therefore we also consider the impact a
heterogeneous population would have on signalling across the network.
EPSRC and Wellcome Trust-MRC Neurodegenerative Disease Initiative
Grant.
References
1. Hardy J, Selkoe DJ: The amyloid hypothesis of Alzheimers disease:
progress and problems on the road to therapeutics. Science 2002,
297(5580):353-356.
2. Walsh D, Klyubin I, Fadeeva JV, Cullen WK, Anwyl R, Wolfe MS, Rowan MJ,
Selkoe DJ: Naturally secreted oligomers of amyloid b protein potently
inhibit hippocampal long-term potentiation in vivo. Nature 2002,
416(6880):535-539.
3. Bliss TV, Collingridge GL: A synaptic model of memory: long-term
potentiation in the hippocampus. Nature 1993, 361(6407):31-39.
4. Destexhe A, Mainen ZF, Sejnowski TJ: An efficient method for computing
synaptic conductances based on a kinetic model of receptor binding.
Neural Comput 1994, 6(1):14-18.
5. Mizuseki K, Diba K, Pastalkova E, Buzski G: Hippocampal CA1 pyramidal
cells form functionally distinct sublayers. Nature Neurosci 2011,
14:1174-1181.
P35
A novel method to find out sensory neuron tracts in the Drosophila
brain
Chao-Chun Chuang
National Center for High-Performance Computing, Taiwan, Republic of China
E-mail: 0203126@narlabs.org.tw
BMC Neuroscience 2015, 16(Suppl 1):P35
How receptions of sensory inputs information turn into perceptions in our
brain? To address these questions, we proposed method reconstructs the
neuronal tracts by applying the shortest path graph algorithm between
functional regions in the Drosophila brain. With these neuronal tracts, we
analyze and draw a network diagram of projection neurons (PNs) relaying
sensory input to higher brain centers in the Drosophila brain. Drosophila is a
widely used genetic model system for understanding human biology [1,2].
While distinctively different in gross anatomy, insect brains and mammalian
brains are both made of neural circuits with a cohort of similar gene
expression governing the basic demands of life. With this network diagram,
numerous unexpected local networks and inter-regional pathways were
found from our initial analysis of sensory systems including olfactory,
gustatory, auditory, and vision circuits [3].
Conclusion: The network diagram shows hierarchical structure, small-world
characteristics, and is composed of functional modules corresponding to the
sensory modalities. This ultimate goal of such an atlas is to identify
connectivity between neurons for understanding the function/circuit
relationships.
References
1. Li CY, Chuang CC, Hua TT, Chen CC, Dickson BJ, Greenspan RJ, et al: A
comprehensive wiring diagram of the protocerebral bridge for visual
information processing in the Drosophila brain. Cell Reports 2013,
3(5):1739-1753.
2. Lee PC, Chuang CC, Chiang AS, Ching YT: High-throughput computer
method for 3d neuronal structure reconstruction from the image stack
of the Drosophila brain and its applications. PLoS Computational Biology
2012, 8(9):e1002658.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P35) Numerous PNs tracts relaying sensory inputs - including olfactory, visual, auditory, and gustatory - to higher brain
centers were discovered
3. Chiang AS, Lin CY, Chuang CC, Chang HM, Hsieh CH, Yeh CW, et al: Three-
dimensional reconstruction of brain-wide wiring networks in Drosophila
at single-cell resolution. Current Biology 2011, 21(1):1-11.
P36
The contribution of subthreshold preference in inhibitory neurons to
network response
Tatjana Tchumatchenko1*, Claudia Clopath2
1 between input patterns with high temporal precision. As input patterns to
Theory of Neural Dynamics Group, Max Planck Institute for Brain Research,
Max-von-Laue-Strasse 4, 60438 Frankfurt, Germany; 2Department of
Bioengineering, Imperial College London, South Kensington Campus,
London SW7 2AZ, UK
E-mail: tatjana.tchumatchenko@brain.mpg.de
BMC Neuroscience 2015, 16(Suppl 1):P36
Oscillations are one of the hallmarks of neural network activity. They have
been implicated in numerous cognitive phenomena and have been
observed in many brain regions. What is interesting about the mechanism
of oscillations in the cortex is that inhibitory neurons seem to play an
important role in the generation and maintenance of these global
oscillations. Yet, is currently unclear what specific properties of these
inhibitory neurons shape the rhythms. One hypothesis is that their intricate
synaptic connections are essential. Alternatively, the intrinsic properties of
the neurons themselves could be the main factor. In the study presented
here and recently published in [1], we explore the combination of both
aspects and consider a network model where interneurons are equipped
with subthreshold resonance and electrical gap junctions. In contrast to
previous studies we address the interaction of both gap junctions and
subthreshold resonance in the same model network. The goal of our work
is to offer a tractable mathematical description of how these two effects
destabilize the fluctuation-driven state of a cortical network. In a cortical
network neurons typically fire irregularly but nevertheless can, under some
conditions, lead to a global resonance or synchronous oscillation. Here, we
show that the oscillation frequency of this rhythm is determined by the
single neuron resonance and modulated by the electrical and chemical
synapses. We argue that the presence of both aspects, gap junctions and
subthreshold resonance, facilitates for the emergence of oscillations in a
cortical network. We find that our results are consistent with several
experimental observations including network responses to oscillatory inputs
in the barrel cortex, see Fig. 1. Thus, they provide a much-needed
computational link connecting different, seemingly disparate effects that
were previously observed in networks.
Reference
1. Tchumatchenko T, Clopath C: Oscillations emerging from noise-driven
steady state in networks with electrical synapses and subthreshold
resonance. Nat Comm 2014, 5:5512.
Page 36 of 200
P37
Extending the tempotron with hierarchical dendrites allows faster learning
Sarah J Jarvis*, Romain Caze, Claudia Clopath
Department of Bioengineering, Imperial College London, SW7 2AZ, UK
E-mail: s.jarvis@imperial.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P37
Certain functional classes of neurons seem to be able to differentiate
neurons can consist of up to thousands of inputs, the ability to identify
target patterns amongst statistically similar background patterns is
impressive and has been suggested to occur via modification to synaptic
weights. Yet how does learning that occurs locally at the level of synapses
converge, without global coordination? While this question is important for
understanding how synaptic learning gives rise to dendritic computation, it
is impractical to test experimentally. Insight from abstract neuronal models,
such as the multilayer perceptron networks [1], provide a potential glimpse
of the difficulty of how to ensure that global convergence during learning
when weight changes are local. In this work, we report that by extending
the tempotron model [2], we are able to demonstrate that learning can,
indeed, learn locally but also converge globally. By arranging dendritic units
in a hierarchical manner which feeds into a master dendrite and soma,
learning occurs over two timescales: locally on each dendritic branch, using
a simple incremental plasticity rule; and at a slower timescale on the main
branch, where information is integrated across branches. We observe that
the inclusion of dendrites reduces the learning time required by allowing
dendrites to subsample the entire input space. In comparison to one single
dendrite receiving n inputs, the inclusion of m dendrites means that each
dendrite is now subsampling n/m inputs, which not results in faster learning
epochs before convergence but also improves the overall robustness
against noise. Thus, the move from a single tempotron to a set of
hierarchically configured tempotrons, representing dendrites, imbues the
unit with recognition of pattern fragments (with a pattern capacity > m!),
faster convergence during learning and increased noise tolerance (both of
which scale with m). The inclusion of dendrites also allows for them to
signal in sequences with varying relative temporal offsets, granting the
neuron the opportunity to differentiate between multiple positive patterns
to identify not only which pattern was observed but also when. Furthermore,
we have also demonstrated that tempotrons can be extended to work for
non-episodic patterns i.e. ongoing and without reset, and can also perform
well when number of distractor patterns greatly outnumber positive
patterns. Conceptually, our model reconciles the tempotron learning rule
with work on dendritic computation which argues for dendrites as
computation units [3]. It also provides a potential explanation to
phenomena observed experimentally, such as neurons in visual cortex
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P36) Neuronal resonance in networks. These results highlight that neuronal models which exhibit subthreshold preference,
regardless of spike generator details, yield results that are consistent with experimental observations in [2]. (A) The response of a recurrent neural
network, RE(2f)/RE(0) to the stimulation of excitatory neurons with periodic stimuli of frequency f (B)The response of a recurrent neural network, RI(2f)/RI
(0), to the stimulation of excitatory neurons with periodic stimuli of frequency f. Figures adapted from [1]
whose dendrites had other preferred orientations that were different to
those of the soma [4].
Acknowledgements: SJ is supported by EU FP7 Marie Curie fellowship
(PIEF-GA-2013-628086). RC is supported by EU FP7 Marie Curie Initial
Training Network 289146 NETT.
References
1. Rumelhart DE, Hinton GE, Williams RJ: Learning representations by back-
propagating errors. Nature 1986, 323:533-536.
2. Gtig R, Sompolinsky H: The tempotron: a neuron that learns spike
timing-based decisions. Nature Neuroscience 2006, 9(3):420-428.
3. Brannon & Mel: Arithmetic of subthreshold synaptic summation in a
model CA1 pyramidal cell. Neuron 2003, 37(6):977-987.
4. Jia H, Rochefort NL, Chen X, Konnerth A: Dendritic organization of sensory
input to cortical neurons in vivo. Nature 2010, 464:1307-1312.
P38
A phase-locked loop epilepsy network emulator for localizing,
forecasting, and controlling ictal activity
Patrick D Watson1,2*, Kevin Horecka1, Rama Ratnam4,5, Neal J Cohen1,3
1
Beckman Institute of Science and Technology, UIUC, IL, USA; 2Neuroscience
Program, UIUC, IL, USA; 3Department of Psychology, UIUC, IL, USA;
4
Coordinated Science Laboratory, UIUC, Urbana, IL, USA; 5Advanced Digital
Sciences Center, Illinois at Singapore Pte. Ltd., Singapore
E-mail: pwatson1@illinois.edu
BMC Neuroscience 2015, 16(Suppl 1):P38
Seizure detection is rapidly improving thanks to novel computational
approaches [1] and public databases of electrocorticography (ECoG) data
[2]. But these approaches rarely model the high-frequency oscillations
that underlie seizure pathology [3]. In the current work, we employ a
phase-locked loop (PLL) neural network [4] to emulate a mesoscale circuit
undergoing high-frequency oscillation. We phase-lock the nodes of the
Page 37 of 200
emulator to raw voltages recorded from chronically implanted ECoG
electrodes in a canine model of epilepsy, and demonstrate that the
emulator experiences uncontrolled phase oscillation far in advance of
either behaviorally observed seizure activity or fluctuations in ECoG
voltages. Using distance weighting to train the epilepsy network
emulator, we localize the ECoG electrodes responsible for destabilization,
and present measures sensitive to these phase disruptions to establish a
forecasting period for ictal activity. We discuss how phase oscillations
from real-time epilepsy network emulation could serve as a closed-loop
feedback control signal to interrupt ictal activity.
References
1. Howbert JJ, Patterson EE, Stead SM, Brinkmann B, Vasoli V, Crepeau D, et al:
Forecasting seizures in dogs with naturally occurring epilepsy. PLoS One
2014, 9(1):e81920.
2. Davis K, Sturges BK, Vite CH, Ruedebusch V, Worrell G, Gardner G, et al: A
novel implanted device to wirelessly record and analyze continuous
intracranial canine EEG. Epilepsy Research 2011, 96(1-2):116-122.
3. Kramer MA, Cash SS: Epilepsy as a Disorder of Cortical Network
Organization. Neuroscientist 2012, 18(4):360-372.
4. Hoppensteadt FC, Izhikevich EM: Pattern recognition via synchronization
in phase-locked loop neural networks. IEEE Transactions on Neural
Networks 2000, 11(3):734-738.
P39
Using phase response curves to predict synchronization times for
neural circuits
Patrick Crotty
Department of Physics and Astronomy, Colgate University, Hamilton,
NY 13346, USA
E-mail: pcrotty@colgate.edu
BMC Neuroscience 2015, 16(Suppl 1):P39
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P39) The mean value of the STDM divided by the period (the combination is dimensionless) as a function of the h and
persistent sodium conductances in the stellate cell, which are the parameters most influencing the intrinsic spiking frequency. The region of
maximal values (red) is approximately in the and low-b region of intrinsic spiking frequency, and also approximately the frequency range where circuits
of simulated cells are observed to synchronize most quickly and where real stellate cells lie. The results shown are for excitatory coupling with synaptic
conductance 0.01 mS / cm2
In a previous work [1], it was found that small simulated circuits of regularly
spiking entorhinal cortex layer II stellate cells (using the model of [2])
synchronize fastest when their intrinsic firing frequencies are approximately
in the 15-20 Hz range, which is very near the frequency range (8-12 Hz)
where these cells are experimentally known to actually fire. The
synchronization time (which we define as the mean time after the onset of
synaptic coupling it takes the cells to synchronize their firings to within one
action potential width of each other, starting with an initially random phase
configuration) in this optimal frequency range can be several times lower
than when the cells have either higher or lower intrinsic frequencies, is
robust across a wide range of 2- and 3-cell circuit topologies and synaptic
coupling strengths, and appears for both excitatory and inhibitory coupling.
The existence of such an optimization may be significant both for the
entorhinal cortex itself (where a background rhythm is believed to play a
role in the phase-coding of position information by grid cells [3]) and in
other parts of the brain for which cell assemblies play an essential role in
information processing, in that assemblies of intrinsically -frequency cells
would be able to form much faster than assemblies of other cells.
The spike-time difference map (STDM) formalism of [2] uses the phase
response curves (PRCs) of two identical coupled cells to predict the
existence and stability of synchronized and other steady-state firing
patterns. The STDM is an iterate of the PRC which gives the amount by
which the time between corresponding spikes in the two cells changes from
one cycle to the next. By taking the mean value of the STDM and dividing
by the period, which gives essentially the mean rate at which the spike
time difference is changing, we find a prominent band of maxima in the
same frequency region as the synchronization time minima (Fig. 1). Like the
synchronization time minimization, the existence and location of this region
of maxima appears to be relatively insensitive to synaptic coupling strength
and excitatory versus inhibitory coupling. Thus, the STDM may provide the
basis for a semi-analytical approach for finding the regions of parameter
space most favorable for synchronization time.
References
1. Crotty P, Lasker E, Cheng S: Constraints on the synchronization of
entorhinal cortex stellate cells. Phys Rev E 2012, 86(Pt 1):011908.
2. Acker CD, Kopell N, White JA: Synchronization of strongly coupled
excitatory neurons: relating network behavior to biophysics. J Comp
Neurosci 2003, 15(1):71-90.
3. Hafting T, Fyhn M, Bonnevie T, Moser MB, Moser E: Hippocampus-independent
phase precession in entorhinal grid cells. Nature 2008, 453:1248-1252.
Page 38 of 200
P40
An asymptotic approximation to the cable equation for arbitrary
diameter taper
Alex D Bird1,2,3*, Hermann Cuntz4,5
1
Warwick Systems Biology Centre, University of Warwick, Coventry, CV4 7AL,
UK; 2School of Life Sciences, University of Warwick, Coventry, CV4 7AL, UK;
3
Warwick Systems Biology DTC, University of Warwick, Coventry, CV4 7AL, UK;
4
Ernst Strngmann Institute for Neuroscience, Frankfurt am Main, Germany;
5
Frankfurt Institute for Advanced Studies, Frankfurt am Main, Germany
E-mail: a.d.bird@warwick.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P40
Somatic integration of synaptic inputs relies on the propagation of currents
arising from sources across the dendritic tree. Whilst active processes
strongly contribute to current flow in most neurons, understanding the
passive backbone to transmission allows a better intuitive grasp of dendritic
function; the results of Rall in highlighting the properties of cylindrical
dendrites[1] are of foundational importance in compartmental modelling
and computational neuroscience. Dendrites are, however, not generally
cylindrical, they tend to taper in a way that contributes to the normalisation
of input currents towards the soma[2].
We have derived an asymptotic approximation to the voltage in dendrites
with an arbitrary taper profile using the insight that voltage attenuation is
substantially faster than radius change in realistic morphologies (Figure 1A).
This result allows faster computation and greater insight than standard
approaches using large numbers of cylinders or frusta to numerically
compute voltage profiles. In addition, it provides easy generalisations of the
standard results of cable theory involving transients and branches.
A particularly interesting implication of these results is that the optimal taper
profile to proximally transmit maximum voltages can be found by variational
calculus, matching results from non-parametric numerical optimisation
which predicted a quadratic form for the diameter taper (Figure 1B) [3].
References
1. Rall W: Branching dendritic trees and motoneuron membrane resistivity.
Exp Neurol 1959, 1(5):491-527.
2. Jaffe DB, Carnevale NT: Passive normalization of synaptic integration
influenced by dendritic architecture. J Neurophysiol 1999, 82(6):3268-3285.
3. Cuntz H, Borst A, Segev I: Optimization principles of dendritic structure.
Theor Biol Med Model 2007, 4:21.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P40) Asymptotic approximations reveal
fundamental properties of voltage flow in tapering dendrites. A
Comparison of numerical (blue) and first-order asymptotic (black
dashed) methods for determining the steady-state voltage in a
quadratically tapering dendritic cable for currents injected at three
different sites. B Comparison of numerically-optimised tapering profiles
(blue) with those predicted by the asymptotic approximation (black
dashed) for dendrites of different electrotonic length
P41
Model-based prediction of maximum pool size in the ribbon synapse
Caitlyn M Parmelee1*, Matthew Van Hook2, Wallace B Thoreson2,3,
Carina Curto4
1
Department of Mathematics, University of Nebraska-Lincoln, Lincoln, NE
68588, USA; 2Department of Ophthalmology and Visual Sciences, University
of Nebraska Medical Center, Omaha, NE 68198, USA; 3Department of
Pharmacology and Experimental Neuroscience, University of Nebraska
Medical Center, Omaha, NE 68198, USA; 4Department of Mathematics, The
Pennsylvania State University, University Park, PA, 16802, USA
E-mail: s-cparmel1@math.unl.edu
BMC Neuroscience 2015, 16(Suppl 1):P41
The synaptic ribbon is a specialized structure in photoreceptor neurons
that tethers vesicles prior to release (Figure 1A). When a cell is stimulated,
vesicles are released from the ribbon and later replenished from the
population of mobile vesicles in the synaptic terminal. A train of
depolarizing pulses causes the ribbon to alternate between periods of
release (lasting t = 25 ms) and replenishment (lasting T = 50ms), which
occur on estimated timescales of r = 5 ms (for release) and a = 815 ms
(for replenishment). After the first few pulses, the system approaches a
limit cycle, and the amount of vesicles released on each pulse converges
to a limiting value, R (Figure 1B). This can be used to determine the
maximum available pool size on the ribbon, A. The standard method for
estimating A is to measure the rate of replenishment in the limit, and then
back-extrapolate from the cumulative release plot to obtain the available
pool size at the start of the pulse train [1]. When comparing pulse trains of
different strengths, this method yields substantially different values for A, a
somewhat paradoxical result. Back-extrapolation assumes, however, that
the replenishment rate is constant, even though it is thought to be
proportional to the available space on the ribbon [2].
Table 1(abstract P41) Maximum pool size predictions from pulse train data
Stimulus Estimate for A, from back-extrapolation
-10 mV (stronger) -136.8794 pA
-30 mV (weaker) -75.1020 pA
Page 39 of 200
We developed a model-based approach to estimate A from the limiting
release R. We modeled the rate of release (resp. replenishment) to simply
be proportional to the number of vesicles on the ribbon (resp. vacant
ribbon sites), and using the measured timescale r (resp. a). By solving
the alternating differential equations, we derived a recurrence relation for
the release during each pulse, R i , which we then solved to obtain a
closed form expression for Ri and the limiting release R. Specifically, we
found that A = cR, where c is a function of r , a ,t,T, and p, with p a
release constant that captures the stimulus dependence of release
probabilities, and can be estimated from the first release, R1. In contrast
to the back-extrapolation method, our model-based estimate for A was
similar across stimulus types (Table 1), while p was much smaller for the
weaker stimulus. This suggests that available pool size does not change
with stimulus strength; instead, differences in release result from changes
in release probability.
References
1. Sakaba T, Schneggenburger R, Neher E: Estimation of quantal parameters
at the calyx of Held synapse. Neurosci Res 2002, 44(4):343-356.
2. Van Hook MJ, Parmelee CM, Chen M, Cork KM, Curto C, Thoreson WB:
Calmodulin enhances ribbon replenishment and shapes filtering of
synaptic transmission by cone photoreceptors. J Gen Physiol 2014,
144(5):357-378.
P42
Coregulation of the Na/K pump and the h-current as a mechanism for
robust neuromodulation
William Barnett1*, Daniel Kueh2, Ronald Calabrese2, Gennady Cymbalyuk1
1
Neuroscience Institute, Georgia State University, Atlanta, GA 30303, USA;
2
Department of Biology, Emory University, Atlanta, GA 30322, USA
E-mail: gcymbalyuk@gsu.edu
BMC Neuroscience 2015, 16(Suppl 1):P42
To achieve the behavioral flexibility necessary for survival in a variable
environment, neuronal circuits have to produce activity over a broad range
of functional characteristics. Central pattern generators (CPGs), rhythmically
active neuronal networks that control motor functions such as breathing,
swimming, and walking, are ideal systems for addressing questions about
how neuromodulation efficiently adjusts network output to meet
environmental demands. Neuromodulators modify the dynamics of CPGs by
orchestrating changes in multiple ionic channels and the electrogenic
pump. The Na+/K+ pump has been implicated as critical in the dynamics of
CPGs [1,2]. In the leech heartbeat CPG, the Na+/K+ pump is a target for
neuromodulation. The neuropeptide myomodulin decreases the burst
period; it increases the h-current and decreases the Na+/K+ pump current.
We developed a model of the leech heart interneuron (HN) including the
Na+/K+ pump current and intracellular Na+ dynamics. We separately
considered the HN and pairs of HNs coupled through mutual inhibition
that form half-center oscillators (HCOs). HCOs form the kernel of the leech
heartbeat CPG. To investigate the role of the h-current and the Na+/K+
pump current, we changed corresponding biophysical parameters of these
currents. We investigated eight model instantiations representing
combinations of three experimental treatments: the blockade of chemical
synapses representing application of bicuculline, the blockade of h-current
representing application of Cs+, and the enhancement of the h-current
and inhibition of the Na+/K+ pump current representing the application of
myomodulin.
Experiments with Cs+ and myomodulin showed the separate effects of
myomodulin on the h-current and the Na+/K+ pump current [1].
Experimental application of myomodulin decreases the period of
oscillatory activity by 17%. Application of Cs+ increases the period of
bursting by 24% relative to control. The application of myomodulin in
addition to Cs+ decreases the period of bursting by 12% relative to
treatment with Cs+. The model captures the qualitative trends in change
Estimate for A, from the model
-131.6858 pA
-133.6100 pA
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P41) A) The synaptic ribbon. (B) The available pool size, A(t), during a stimulus pulse train
of cycle period observed in experiments with myomodulin and Cs+. Since
the application of myomodulin decreases the period of activity when the
h-current is present or absent, the Na+/K+ pump current plays a significant
role in the dynamics of the HCO.
We investigated activity of the model with different values of the
maximal conductance of the h-current, g h and the maximal pump
current, IMax
Pump , for isolated neurons and HCOs. We found ranges of
parameters where neurons showed bursting activity. In the HCO, we
identified the region of the parameter space that supported functional
bursting activity. Joint changes of IMax
Pump and g h allows neurons to
preserve functional activity over a larger span of IMax
Pump and g h and to
produce functional activity with a larger range of period.
Acknowledgements: Supported by NINDS 1 R01 NS085006 to RLC and
by NSF PHY-0750456 to GSC.
References
1. Tobin AE, Calabrese RL: Myomodulin increases Ih and inhibits the NA/K
pump to modulate bursting in leech heart interneurons. J Neurophysiol
2005, 94(6):3938-3950.
2. Zhang HY, Sillar KT: Short-term memory of motor network performance
via activity-dependent potentiation of Na+/K+ pump function. Curr Biol
2012, 22(6):526-531.
P43
Robustness of spatial learning in flickering networks
Yuri A Dabaghian1,2*, Samir Chowdhury3, Andrey Babichev1,2,
Facundo Mmoli3
1
Neurology-Pediatrics Department, Baylor College of Medicine, Houston, TX
77030, USA; 2Computational and Applied Mathematics, Rice University,
Houston, TX, 77005, USA; 3Department of Mathematics, Ohio State
University, Columbus, OH, 43210, USA
E-mail: dabaghia@bcm.edu
BMC Neuroscience 2015, 16(Suppl 1):P43
It is widely accepted that the network of the hippocampal place cells
provides a substrate of the cognitive map of the environment. However,
thousands of hippocampal neurons die every day and the networks
formed by these cells constantly change due to various forms of synaptic
plasticity. What then explains the remarkable reliability of our spatial
memories? We propose a computational approach to answering this
question based on a couple of insights. First, we propose that the
hippocampal cognitive map is fundamentally topological, i.e., more similar
to a subway map than to a topographical city map [1], and hence it is
amenable to analysis by topological methods [2,3]. We then apply several
novel methods from homology theory, to understand how dynamic
connections between cells influences the speed and reliability of spatial
learning. We simulate the rats exploratory movements through different
environments and study how topological invariants (stable topological
features of different experimental environments) arise in a network of
Page 40 of 200
simulated neurons with dynamic, flickering connectivity. We find that
despite transient connectivity the network of place cells produces a stable
representation of the topology of the environment.
Acknowledgements: Work is supported by the NSF grant NSF 1422438.
References
1. Dabaghian Y, Brandt V, Frank L: Reconceiving the hippocampal map as a
topological template. eLife 2014, 3:e03476.
2. Arai M, Brandt V, Dabaghian Y: The Effects of Theta Precession on Spatial
Learning and Simplicial Complex Dynamics in a Topological Model of
the Hippocampal Spatial Map. PLoS Comput Biol 2014, 10(6):e1003651.
3. Dabaghian Y, Mmoli F, Frank L, Carlsson G: A Topological Paradigm for
Hippocampal Spatial Map Formation Using Persistent Homology. PLoS
Comput Biol 2012, 8(8):e1002581.
P44
A topological approach to synaptic connectivity and spatial memory
Russell Milton1, Andrey Babichev2,3, Yuri A Dabaghian2,3*
1
Department of Neurobiology and Anatomy, University of Texas Health Science
Center at Houston, TX 77030, USA; 2Neurology-Pediatrics Department, Baylor
College of Medicine, Houston, TX 77030, USA; 3Computational and Applied
Mathematics, Rice University, Houston, TX 77005, USA
E-mail: dabaghia@bcm.edu
BMC Neuroscience 2015, 16(Suppl 1):P44
In the hippocampus, a network of place cells generates a cognitive map of
space, in which each cell is responsive to a particular area of the
environment its place field. The peak response of each cell and the size of
each place field have considerable variability. Experimental evidence
suggests that place cells encode a topological map of space that serves as a
basis of spatial memory and spatial awareness. Using a computational
model based on Persistent Homology Theory we demonstrate that if the
parameters of the place cells spiking activity fall inside of the physiological
range, the network correctly encodes the topological features of the
environment. We next introduce parameters of synaptic connectivity into
the model and demonstrate that failures in synapses that detect coincident
neuronal activity lead to spatial learning deficiencies similar to the ones that
are observed in rodent models of neurodegenerative diseases. Moreover, we
show that these learning deficiencies may be mitigated by increasing the
number of active cells and/or by increasing their firing rate, suggesting the
existence of a compensatory mechanism inherent to the cognitive map.
Acknowledgements: Work is supported by the NSF grant NSF 1422438.
References
1. Dabaghian Y, Brandt V, Frank L: Reconceiving the hippocampal map as a
topological template. eLife 2014, 3:e03476.
2. Arai M, Brandt V, Dabaghian Y: The Effects of Theta Precession on Spatial
Learning and Simplicial Complex Dynamics in a Topological Model of
the Hippocampal Spatial Map. PLoS Comput Biol 2014, 10(6):e1003651.
3. Dabaghian Y, Mmoli F, Frank L, Carlsson G: A Topological Paradigm for
Hippocampal Spatial Map Formation Using Persistent Homology. PLoS
Comput Biol 2012, 8(8):e1002581.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
P45
Towards biophysical psychiatry": a modeling approach for studying
effects of schizophrenia-linked genes on single-neuron excitability
Tuomo Mki-Marttunen1*, Geir Halnes2, Anna Devor3,4, Aree Witoelar1,
Francesco Bettella1, Srdjan Djurovic5, Yunpeng Wang3, Gaute T Einevoll2,
Ole A Andreassen1, Anders M Dale3,4
1
NORMENT, Institute of Clinical Medicine, University of Oslo, Oslo, Norway;
2
Department of Mathematical Sciences and Technology, Norwegian
University of Life Sciences, s, Norway; 3Department of Neurosciences,
University of California San Diego, La Jolla, San Diego, CA, USA; 4Department
of Radiology, University of California San Diego, La Jolla, San Diego, CA, USA;
5
Department of Medical Genetics, Oslo University Hospital, Oslo, Norway
E-mail: tuomomm@uio.no
BMC Neuroscience 2015, 16(Suppl 1):P45
Genome-wide association studies (GWAS) employing large sample sizes
and sophisticated statistical methods have recently yielded detailed
information on the set of genes affected in various psychiatric disorders.
This is especially important for polygenic, highly heritable disorders such
as schizophrenia (SCZ) [1]. The success lately witnessed in gene discovery
brings up the next big challenge for psychiatric genetics - translation of
the genetic associations into biological insights [2]. In this study, we
propose a computational approach for investigating the effects of a
collection of excitability-related genes on various neuronal characteristics.
As a proof of principle, we apply our approach for studying effects of SCZ-
linked genes on firing behavior of a layer V pyramidal cell (L5PC).
A total of 108 genetic loci were recently confirmed to be associated with the
risk of SCZ [3]. These loci span a wide set of protein-coding genes. The
disorder is associated with genes affecting transmembrane currents of all
major cationic species, Na+, K+, and Ca2+. In addition, some of the SCZ-
linked genes are involved in regulation of the Ca2+ concentration in the
intracellular medium, which is another great contributor to neuron
excitability. All above aspects of electrophysiology are included in a recent
multi-compartmental model of L5PC [4], which accurately describes the
perisomatic firing behavior and its interplay with the generation of an apical
Ca2+ spike. In this work, we rely on data from functional genomics studies
that describe the effects of variants of certain ion channel or calcium
transporter-encoding genes on the channel activation/inactivation
properties or intracellular Ca2+ dynamics. We carry out our study by linking
these effects to a change in the corresponding neuron model parameters,
and observing the implication that these variants have on the information
integration in an L5PC.
It should be noted that information does not generally exist for the effect
of single nucleotide polymorphism (SNP) variants identified through
GWASs on the biophysical parameters required for the computational
models. We instead use information obtained from in vitro studies of more
extreme genetic variations, including loss of function mutations. A central
assumption of this approach is that the effects of SNP variants can be
represented as scaled-down versions of those of the more extreme
variants, and that the emergence of the full psychiatric disease phenotype
results from the combined effect of a large number of subtle SNP effects.
Our results show a multitude of alterations of the firing behavior and
integration of apical stimuli in neurons equipped with the considered
gene variants. An especially interesting observation is that the variants
affect the suppression of a second apical stimulus, which might have a
connection with the deficit in prepulse inhibition, a condition often
observed in SCZ patients. Although the analyses presented here are
specific to L5PCs and SCZ-related genes, our biophysical psychiatry
framework may be directly applicable to other cell types and other
polygenic diseases, such as bipolar disorder and autism, given an
identification of risk genes related to neuronal excitability. Furthermore,
our approach could be directly applied to biophysically detailed models
of neuronal networks and extended to consider synaptic ion channel-
encoding genes that are relevant in SCZ [5].
References
1. Ripke S, Sanders AR, Kendler KS, Levinson DF, Sklar P, Holmans PA, et al:
Genome-wide association study identifies five new schizophrenia loci.
Nat Gen 2011, 43(10):969-976.
2. van Os J, Kapur S: Schizophrenia. Lancet 2009, 374(9690):635-645.
Page 41 of 200
3. Ripke S, Neale BM, Corvin A, Walters JT, Farh KH, Holmans P, et al:
Biological insights from 108 schizophrenia-associated genetic loci. Nature
2014, 511:421-427.
4. Hay E, Hill S, Schrmann F, Markram H, Segev I: Models of neocortical
layer 5b pyramidal cells capturing a wide range of dendritic and
perisomatic active properties. PLoS Comput Biol 2011, 7(7):e1002107.
5. Wen Z, Nguyen HN, Guo Z, Lalli MA, Wang X, Su Y, et al: Synaptic
dysregulation in a human iPS cell model of mental disorders. Nature
2014, 515(7527):414-418.
P46
Nonlinear system identification of receptive fields from spiking neuron
data
Dorian Florescu, Daniel Coca*
Department of Automatic Control and Systems Engineering, University of
Sheffield, Sheffield, S1 3JD, UK
E-mail: d.coca@sheffield.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P46
Identification of a linear filter in cascade with a spiking neuron has been
previously considered [1] under the assumption that the input of the
Hodgkin-Huxley spiking neuron can be measured in addition to the input
and output of the circuit. An alternative method to identify a linear filter
given the input to the circuit and the spike time sequence, has been
proposed [2] for Integrate-and-Fire (IAF) neurons. The method was further
extended [3] to nonlinear receptive fields that can be described by a
Volterra series.
Here we propose a new system identification methodology to identify a
more general NARMAX representation of the nonlinear receptive field in
cascade with an ideal-IAF neuron, based only on measurements of the
input stimulus and the spike time sequence of the IAF neuron. By using an
orthogonal forward selection algorithm we are able to derive the NARMAX
representation of a scaled version of the nonlinear filter directly from the
reconstructed input of an ideal IAF neuron. The method is further
extended to leaky-IAF neurons which require estimating an additional
spiking neuron parameter. We use the NARMAX methodology to identify
recursively the nonlinear filter and the spiking neuron parameter in
the presence of noise. Statistical and dynamical model validation tests are
used to check if the identified nonlinear filter models are an adequate
representation of the underlying nonlinear information processing
mechanisms. The performance and robustness to noise of the proposed
methods is demonstrated through numerical simulation studies.
Specifically, we compared the generalized (higher-order) frequency
response functions (GFRFs) of the original and the identified nonlinear
filter (Figures 1 A-D) and evaluated, for different noise levels, the SNR of
the model predicted output on a validation data set (Figure 1 E).
References
1. Lazar AA, Slutskiy YB: Functional identification of spike-processing neural
circuits. Neural Computation 2014, 26(2):264-305.
2. Lazar AA, Slutskiy YB: Identifying dendritic processing. Adv Neural Inf
Process Syst 2010, 23:1261-1269.
3. Lazar AA, Slutskiy YB: Spiking neural circuits with dendritic stimulus
processors: encoding, decoding, and identification in Reproducing
Kernel Hilbert Spaces. Journal of Computational Neuroscience 2015,
38(1):1-24.
P47
An empirical model of Drosophila Photoreceptor-LMC network
Carlos Luna, Daniel Coca*
Department of Automatic Control and Systems Engineering, University of
Sheffield, Sheffield, S101EQ, UK
E-mail: d.coca@sheffield.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P47
To operate over the full environmental range of light intensities, fly
photoreceptors implement various adaptation mechanisms to continuously
adjust their sensitivity. In addition, photoreceptor responses in flies are
modulated by feedback from two classes of interneuron, large monopolar
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 42 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

Figure 1(abstract P46) A&B - GFRF functions of the original filter, C&D - Difference between the original GFRFs and the scaled versions for the
identified filter, E - SNR between the validation signal and the identified filter output

Figure 1(abstract P47) A. Block diagram of a single photoreceptor-LMC circuit model. B. Experimental recording (black) vs model predicted output (red)

cells (LMC) and amacrine cells (AC), and axonal gap-junctions, which pool
the responses from six photoreceptors. Previous studies have shown that
adaptation in the photoreceptor-interneuron synapses helps extend further
the operating range of photoreceptors.
In histamine deficient mutants hdc JK910 the lamina interneurons fail to
receive and transmit visual information and their feedback synapses can
no longer modulate the photoreceptor output. By comparing the open-
loop system of mutant flies with the closed-loop system of wild-type flies
it is possible to characterize how the network helps enhance the
photoreceptor response. We have previously developed a functional
nonlinear model of Drosophilas R1-R6 which incorporates separate gains
for mean intensity and contrast. By exploiting this control architecture we
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
were able to model the responses of wild type as well as blind mutants
using a single model structure and to characterize the contribution from
interneurons to boosting the operating ranges of fly photoreceptors.
Here we propose a new control structure (see Figure 1 panel A), which
separates the gain control implemented by individual photoreceptors from
the photoreceptor-LMC feedback-control loop. By exploiting this control
architecture we were able to develop an empirical model of Drosophila
retinal network circuit, which couples six photoreceptor models (R1-R6)
with a model of the LMCs. The model not only predicts the individual
photoreceptor responses of wild type but also those of hdcJK910 mutants
by simply disconnecting the LMC feedback loop (see Figure1 Panel B). We
used the model to demonstrate the potential role of the retinal circuit in
increasing the Signal-to-Noise Ratio of the incoming light stimuli. The
elementary circuit model, which was validated extensively using
experimental data, can be used to implement a complete spatio-temporal
model of the flys retina.
P48
Modeling of respiratory network: to sigh or not to sigh
Tatiana Dashevskiy1*, Jan-Marino Ramirez1,2
1
Center for Integrative Brain Research, Seattle Childrens Research Institute,
Seattle, WA, USA; 2Department of Neurological Surgery, University of
Washington, Seattle, WA, USA
E-mail: tatiana.dashevskiy@seattlechildrens.org
BMC Neuroscience 2015, 16(Suppl 1):P48
The Pre-Btzinger Complex (PreBtC), a medullary brainstem region, is
essential for generating breathing. Isolated in a transverse slice
preparation, the preBtC continues to generate under normal conditions
fictive eupneic activity and at a lower frequency also fictive sighs. The in some sense, easier from tripod to tetrapod than in the opposite direction.
mechanisms generating the periodicity of fictive sighs as well as its
intrinsic nature are largely unknown. Based on the experimental
observations such as 1) sigh has the biphasic shape characterized as
eupneic breath interrupted by augmentation with a second peak of
activity, 2) the biphasic shape persists after blockage of inhibitory
synapses, 3) the network is heterogeneous, we model PreBtC network
that reproduces both types of activities. The model contains three types of
neurons: intrinsically tonic spiking, intrinsic bursting neurons and intrinsic
quiescent. Neurons within the network are sparsely connected through
randomly distributed excitatory synapses. In this model we propose that
the mechanism of sigh generation is based on slow oscillations generated
by glia. Moreover, the model predicts that depending on the modularity
state of the system, the network exhibits sighs, does not exhibit sighs and
shows only eupneic activity, or shows bi-stable activity meaning that sighs
can be triggered between on and off states by short perturbation. The
bistability depends on slow calcium oscillations. The network model
predicts that the presence of IP3-like channels is necessary to produce
sigh behavior and to mimic transitions between activity states caused by
neuromodulators. We analyze the model dynamics and describe dose-
response curve to neuromodulator such as norepinephrine. Theoretical
data are similar to experimental data from in vitro brain slice preparations
and in vivo anesthetized mice.
P49
A model postulating a pivotal role of the levator-depressor neuro-
muscular systems in locomotion of the stick insect
Tibor I Toth*, Silvia Daun-Gruhn
Heisenberg Research Group of Computational Biology, Institute of Zoology,
University of Cologne, 50674 Cologne, Germany
E-mail: ttoth0@uni-koeln.de
BMC Neuroscience 2015, 16(Suppl 1):P49
Locomotion of stick insects has been intensively studied for many years.
These studies made use of both behavioural [1,2] and electrophysiological
e.g. [3,4] methods. In addition, models have been constructed to mimic
the locomotion (normal walking) of these insects e.g. [5]. All these studies
have substantially advanced our knowledge on insect locomotion.
Nevertheless, there still exist a large number of unresolved problems
in this field. One of them is about the possibly hierarchical organization
of the coordinated movements of the leg-joints during normal walking.
Page 43 of 200
The existing experimental data relating to this problem have not proved
decisive. A modelling approach would therefore seem particularly apt to
contribute to the resolution of this problem. The advantage of this method
is that a wide range of scenarios and hypothetical experimental
arrangements can be emulated by a model that bears the main hallmarks
of its biological counterpart.
We followed exactly this route when investigating the organization of
locomotion. We thus extended an earlier model of ours [6] that emulates
the coordinated muscle activities in a single (middle) leg of the stick insect
to include all three ipsilateral legs. The model of each leg consisted of the
three main neuro-muscular systems that correspond to the three main
antagonistic muscle pairs: m. protractor and retractor coxae, m. levator and
depressor trochanteris and m. extensor and flexor tibiae. The muscles were
activated by corresponding motoneurones which, in turn, were controlled
by the activity of non-spiking pre-motor interneurones and the activity of
central pattern generators (CPGs). Our basic assumption was that the
levator-depressor systems of each leg play a pivotal role in the
organization of the coordinated movement of the three ipsilateral legs.
According to our reasoning, the most reliable signal that affects normal
stepping during locomotion is ground contact (touch-down) and lift-off.
Both of them are, in essence, brought about by vertical movement of the
femur of the leg, that is by proper activation of the levator-depressor
neuro-muscular system.
We tested this assumption accordingly in simulations with the model. In
doing so, we succeeded in mimicking both the tetrapod and the tripod
coordination patterns, which stick insects exhibit during normal walking.
Moreover, we found that the coordination process in tripod only requires
position signals of the legs whereas in tetrapod angular velocity signals, too,
are necessary for the coordination. We also succeeded in modelling the
transition between these coordination patterns and found that the switch is,
The model thus supports our assumption on the pivotal role of the levator-
depressor neuro-muscular system in normal walking.
Acknowledgements: This work has been funded by the Deutsche
Forschungsgemeinschaft Grants DA1182/1-1 and GR3690/4-1.
References
1. Cruse H: What mechanisms coordinate leg movement in walking
arthropods? TINS 1990, 13:15-21.
2. Graham D: A behavioural analysis of the temporal organisation of
walking movements in the 1st instar and adult stick insect (Carausius
morosus). J Comp Physiol A 1972, 81:23-52.
3. Borgmann A, Scharstein H, Bschges A: Intersegmental coordination:
influence of a single walking leg on the neighboring segments in the
stick insect walking system. J Neurophysiol 2007, 98:1685-1696.
4. Borgmann A, Hooper SL, Bschges A: Sensory feedback induced by front-
leg stepping entrains the activity of central pattern generators in caudal
segments of the stick insect walking system. J Neurosci 2009,
29:2972-2983.
5. Cruse H, Kindermann T, Schumm M, Dean J, Schmitz J: Walknet - a
biologically inspired network to control six-legged walking. Neural
Networks 1998, 11:1435-1447.
6. Knops S, Toth TI, Guschlbauer C, Gruhn M, Daun-Gruhn S: A
neuromechanical model for the neuronal basis of curve walking in the
stick insect. J Neurophysiol 2013, 109:679-691.
P50
Modelling searching movements of the front leg in the stick insect by
means of a neuro-muscular model
Tibor I Toth1*, Eva Berg2, Ansgar Bschges2, Joachim Schmidt2,
Silvia Daun-Gruhn1
1
Heisenberg Research Group of Computational Biology, Institute of Zoology,
University of Cologne, 50674 Cologne, Germany; 2Department of Animal
Physiology, Institute of Zoology, University of Cologne, 50674 Cologne,
Germany
E-mail: ttoth0@uni-koeln.de
BMC Neuroscience 2015, 16(Suppl 1):P50
Searching movements, beside locomotion, are, perhaps, the most important
motor activities of the front legs in stick insects. In a recent study [1], the
kinematics of these movements were thoroughly investigated. In these
experiments the animal was restricted such that its front leg could freely
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
move in the plane perpendicular to the longitudinal axis of the animal, only.
Having induced leg movements in the stick insect, an obstacle was put in
the way of the distal tibia at different angle positions with respect to the
horizontal axis. The obstacle was however removed immediately after
collision with the insects leg. Leg movements before and after the animals
touching of the obstacle were recorded and analyzed. In particular, the
angle between the horizontal axis and the femur (b), and that between
the femur and tibia (g) were measured. In addition, the angle (a) describing
the position of the distal tibia was also determined. It was found that the
front leg after collision with the obstacle showed modified searching
movements that could be regarded as targeted response expressed by the
time courses of the angles a, b and g. The specific properties of the targeted
searching movements depended on the position of the obstacle.
Using our existing model that was constructed to emulate stepping of a
single middle leg [2], we tried to mimic the aforementioned experiments.
First, we replaced the anatomical data of the middle leg with those of the
front leg of the stick insect. We then fixed the protractor-retractor angle at
90 in the model, as it was done in the experiments. We implemented the
searching movements by appropriate activation or deactivation of fast and
slow muscle groups of the levator-depressor and the extensor-flexor
neuro-muscular system. At touch of an obstacle, all muscle activity was
stopped for a very short time, and the recruitment of the muscles in the
individual muscle groups was also reduced to a minimum. Restoring the
activity and recruitment in the individual muscle group took place
depending on the position of the obstacle. We thus succeeded in
replicating the main properties of the leg movements before and after its
touching the obstacle. We found that the time course of the recruitment of
the fibres in the fast levator and depressor muscles crucially affected the
time courses of the angles a, b and g, hence the overall (2-dimensional)
movement of the leg. The threshold value of the angle b at which the
extension switches to flexion and vice versa in the extensor-flexor neuro-
muscular system has turned out to be another important system
parameter. At lowering the value of this parameter, the time course of g
becomes increasingly erratic, and the tibia eventually stops oscillating.
Such events could also be observed in the experiments.
The good agreement between experimental and simulation results suggests
that the properties and parameters used in the model to mimic search
behaviour, might also be those that affect the searching movements in the
stick insect in an analogous way.
Acknowledgements: This work has been funded by the Deutsche
Forschungsgemeinschaft Grants DA1182/1-1 and GR3690/4-1.
References
1. Berg E, Bschges A, Schmidt J: Single perturbations cause sustained
changes in searching behavior in stick insects. J Exp Biol 2013,
216:1064-1074.
2. Toth TI, Grabowska M, Schmidt J, Bschges A, Daun-Gruhn S: A neuro-
mechanical model explaining the physiological role of fast and slow
muscle fibres at stop and start of stepping of an insect leg. PLoS One
2013, 8(11):e78246.
P51
Invariance to frequency and time dilation along the ascending ferret
auditory system
Alexander G Dimitrov1*, Jean F Lienard1, Zachary Schwartz2, Stephen V David2
1 the environment in which they are embedded [1]. In this approach, large
Department of Mathematics, Washington State University Vancouver,
Vancouver, WA 98686, USA; 2Department of Otolaryngology, Oregon Health
Sciences University, Portland, OR 97239, USA
E-mail: alex.dimitrov@vancouver.wsu.edu
BMC Neuroscience 2015, 16(Suppl 1):P51
The sense of hearing requires a balance between competing processes of
perceiving and ignoring. Behavioral meaning depends on the combined
values of some sound features but remains invariant to others. The
invariance of perception to physical transformations of sound can be
attributed in some cases to local, hard-wired circuits in peripheral brain
areas. However, at a higher level this process is dynamic and continuously
adapting to new contexts throughout life. Thus the rules defining invariant
features can change.
In this project, we test the idea that high-level, coherent auditory
processing is achieved through hierarchical bottom-up combinations of
neural elements that are only locally invariant. The main questions we
Page 44 of 200
address in the context of an auditory system are: 1. What kinds of changes
in sound do not affect initial stages of auditory processing? 2. How does
the brain manipulate these small effects to achieve a coherent percept of
sounds?
Local probabilistic invariances, defined by the distribution of transformations
that can be applied to a sensory stimulus without affecting the
corresponding neural response [1,2], are largely unstudied in auditory
cortex. We assess these invariances at two stages of the auditory hierarchy
using single neuron recordings from the primary auditory cortex (A1) and
the secondary auditory cortex (PEG) of awake, passively listening ferrets [3,4].
Our results show that stimulus invariance to frequency and time dilations
are present at every tested stage and increase along the hierarchical
auditory processing. At least in the early stages, parametric models having
invariance properties by design are well-suited to describing biological
functions. We were further able to characterize meaningful relationships
among receptive field shapes. Preliminary observations indicate that joint
time/frequency receptive fields are oriented toward central frequencies;
receptive field widths are proportional to the best frequency; and late-
onset neurons are also exhibiting the most sustained activity.
References
1. Aldworth ZN, Miller JP, Gedeon T, Cummins GI, Dimitrov AG: Dejittered
spike-conditioned stimulus waveforms yield improved estimates of
neuronal feature sensitivity. J Neuro 2005, 25(22):5323-5332.
2. Dimitrov AG, Gedeon T: Effects of stimulus transformations on the
perceived function of sensory neurons. J Comp Neuro 2006, 20:265-283.
3. Atiani S, David SV, Elgueda D, Locastro M, Radtke-Schuller S, Shamma SA,
Fritz JB: Emergent selectivity for task-relevant stimuli in higher order
auditory cortex. Neuron 2014, 82(2):486-99.
4. David SV, Mesgarani N, Shamma SA: Estimating sparse spectro-temporal
receptive fields with natural stimuli. Network 2007, 18(3):191-212.
P52
Purkinje cells: the forest shapes the trees
Benjamin Torben-Nielsen*, Erik De Schutter
Computational Neuroscience Unit, Okinawa Institute of Science and
Technology, 1919-1 Tancha, Onna-son, Kunigami-gun, Okinawa, 904-0495,
Japan
E-mail: b.torbennielsen@gmail.com
BMC Neuroscience 2015, 16(Suppl 1):P52
Purkinje cells are at the heart of all theories about the olivo-cerbellar circuit
as they provide the only output from the cerebellar cortex. Hallmark is
their planar morphology in a plane perpendicular to the parallel fibers.
Moreover, individual Purkinje cells have very dense dendritic arborizations,
a feature commonly referred to as space-filling. In a population, they
neatly align and cover the available volume in both transverse and medio-
lateral directions, so-called tiling. These remarkable morphological features
are assumed to allow neurons to intercept as many parallel fibers inputs as
possible; up to 250.000 in individual cells. But it remains an open question
how individual cells develop a mono-planar dendrite and how the
alignment emerges in a population. Here we address both questions.
We recently developed a computational framework, NeuroMaC, to
investigate how neuronal morphologies are shaped by interactions with
numbers of virtual morphologies are generated simultaneously by
repeatedly extending simulated growth-cones that follow growth rules
expressed in terms of environmental interactions. In this work, we
hypothesized that the peculiar Purkinje cell morphology emerges from
three simple interactions with its environment. First, dendritic branches of
one neuron are repelled by each other, a process called self-avoidance.
Second, dendritic branches of different neurons repel each other. Third, all
branches are to some extent attracted to the pia surface. Initial simulated
growth-cones where placed on a seven-by-seven lattice with plausible
spacing: somata where 25 and 150 micron apart from each other in,
respectively, the medio-lateral and transverse directions. We found that
these simple rules yielded highly realistic morphologies exhibiting the
hallmark planarity in the transverse direction. The planarity emerged
because the repulsion between branches forced the neuron to occupy all
available space, which was larger in the transverse than in the medio-
lateral direction. We validated the resulting morphologies against exemplar
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P51) 95% confidence interval in frequency for the highest responses in A1 (A) and PEG (B) as tested with narrowband noise
stimuli. The higher values found in PEG compared to A1 demonstrate higher invariance to frequency shifts in PEG than in A1for both stimuli. (C) A ferret
vocalization (top) and three responses to different time dilations (bottom). The inset shows the aligned firing rates, indicating similarities and differences
in the responses
data and found a good statistical match. Moreover, alignment and tiling
emerged in the population.
We demonstrated that key features of the notoriously complex Purkinje cell
morphology could be achieved by solely local repulsive and attractive
interactions between simulated growth-cones. Thus, we may conclude that
the forest does shape individual neurons. While alignment did emerge to a
certain extent, it is hard to validate because there is currently no data
available on Purkinje cell forests. We also speculate on the necessity of a
global organization mechanism that consistently aligns Purkinje cells in the
transverse direction.
Reference
1. Torben-Nielsen B, De Schutter E: Context-aware modeling of neuronal
morphologies. Front Neuroanat 2014, 8:92.
P53
The ionic mechanism of the Purkinje cell dendritic spikes generation
and propagation: a model exploration
Yunliang Zang*, Erik De Schutter
Computational Neuroscience Unit, Okinawa Institute of Science and
Technology Graduate University, Onna-son, Okinawa 904-0895, Japan
E-mail: erik@oist.jp
BMC Neuroscience 2015, 16(Suppl 1):P53
Due to the lack of sodium channel expression in the Purkinje cell (PC)
dendrite, the passively propagated sodium spikes can hardly be detected
in the dendrite. However, the substantial expression of T type calcium
channels (I CaT ) and P type calcium channels (I CaP ) can cause local
regenerative spikes to occur in the dendrite after receiving a parallel fiber
(PF) stimulus [1].
In this study, with the previously published PC dendrite model [2] from
our group as the starting point, we have built a new PC dendrite model
incorporating the low threshold A-type K+ current (IKA) [3], high threshold
activated K+ current (IKV3) and hyperpolarization-activated cation current
(Ih). In our model, in agreement with experimental observations [1], the
EPSP amplitude increases nonlinearly with the strength of the PF synaptic
input until it reaches the spike threshold. During this range, we find that
ICaT amplifies the EPSP caused by PF synaptic input to boost the EPSP
close to the spike threshold. Around the spike threshold there is an
obvious spike delay due to the interactive role of IKA and ICaT. Above the
spike threshold, high threshold activated ICaP takes over and depolarizes
the dendrite rapidly, finally leading to the generation of dendritic spikes.
The amplitude of dendritic spikes doesnt increase further with the
stimulation intensity and is mainly determined by the struggle among
I CaP , I KV3 and high conductance calcium activated K + currents (I BK ).
Injecting a negative current can eliminate the PF input evoked dendritic
spikes by lowering the membrane potential which implies less activation
Page 45 of 200
of the ICaT and subsequent less amplification of the EPSP. We also find
that the simulated dendritic spikes cant propagate to the whole dendrite
due to the regulating role of mainly IBK and IKV3. Surprisingly by partially
blocking IKV3 and IBK parallel fiber evoked dendritic spikes can propagate
to the whole dendrite because of the broadening of its width.
The calcium influx by dendritic spikes is vital to a lot of calcium dependent
processes like synaptic plasticity. Understanding the mechanism of
generation of the local dendritic spikes evoked by PF stimulus and how
their localized nature works will help us better explore how information is
processed in the PC dendrite.
References
1. Rancz EA, Hausser M: Dendritic calcium spikes are tunable triggers of
cannabinoid release and short-term synaptic plasticity in cerebellar
Purkinje neurons. J Neurosci 2006, 26(20):5428-5437.
2. Anwar H, Hong S, De Schutter E: Controlling Ca2+-activated K+ channels
with models of Ca2+ buffering in Purkinje cells. Cerebellum 2012,
11(3):681-693.
3. Otsu Y, Marcaggi P, Feltz A, Isope P, Kollo M, Nusser Z, et al: Activity-
dependent gating of calcium spikes by A-type K+ channels controls
climbing fiber signaling in Purkinje cell dendrites. Neuron 2014,
84(1):137-151.
P54
Implementation of parallel spatial stochastic reaction-diffusion
simulation in STEPS
Weiliang Chen1*, Iain Hepburn1,2, Erik De Schutter1,2
1
Computational Neuroscience Unit, Okinawa Institute of Science and
Technology, Okinawa 904-0411, Japan; 2Theoretical Neurobiology, University
of Antwerp, B-2610 Antwerpen, Belgium
E-mail: erik@oist.jp
BMC Neuroscience 2015, 16(Suppl 1):P54
Spatial stochastic reaction-diffusion simulation has been recognized as an
essential modeling tool in computational neuroscience studies of signaling
pathways, as shown in an increasing number of recent studies such as our
previous work on the stochastic effects of calcium dynamics in Purkinje
cells [1]. A critical performance issue arises when attempting to model
large pathway models with complex morphologies, for example the one
proposed in Human Brain Project [2], due to the serial nature of Gillespies
direct method [3], the fundamental algorithm of many spatial stochastic
reaction-diffusion simulators including STEPS [4]. Various solutions have
been proposed to improve the computational efficiency of the Gillespie
method, including the tau-leaping approximation [5], most of which,
however, remain serial implementations.
The need of parallel implementation of stochastic reaction-diffusion
simulators has become urgent, as the scale and complexity of model
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
being studied surpasses the speedup gained from hardware upgrade and
algorithm improvement. However, such a task is not trivial as the original
Gillespie SSA is known to be extremely serial.
In CNS2014 we proposed a parallel solution to approximate diffusion
events in STEPS [6], which significantly improves the performance while
maintaining high accuracy. We now further improve this solution by
introducing a multinomial algorithm for fast diffusion direction selection of
multiple molecules in a single subvolume. We also combine this diffusion
approximation with a new operator splitting solution for reaction events in
the SSA system. The combined solution is implemented in STEPS as its first
parallel solver named TetOpSplit. Current implementation of TetOpSplit
uses MPI as its parallel protocol and aims to provide solutions for large
scale simulations such as whole cell reaction-diffusion models, in modern
supercomputers like Blue Gene.
In this poster we discuss the difficulties we encountered during the
transformation from the serial TetOpSplit algorithm to its parallel
counterpart, as well as our solutions. We also provide performance results
of the new solver via different examples, and compare them with the
results gained from our original serial SSA implementation.
References
1. Anwar H, Hepburn I, Nedelescu H, Chen W, De Schutter E: Stochastic
Calcium Mechanisms Cause Dendritic Calcium Spike Variability.
J Neurosci 2013, 33(40):15488-15867.
2. The Human Brain Project. [http://www.humanbrainproject.eu].
3. Gillespie DT: Exact Stochastic Simulation of Coupled Chemical Reactions.
The Journal of Physical Chemistry 1977, 81(25):2340-2361.
4. Hepburn I, Chen W, Wils S, De Schutter E: STEPS: efficient simulation of
stochastic reaction-diffusion models in realistic geometries. BMC Syst Biol
2012, 6:36.
5. Gillespie DT: Approximate accelerated stochastic simulation of chemically
reacting systems. The Journal of Chemical Physics 2001, 115(4):1716-1711.
6. Hepburn I, et al: Accurate approximation and MPI parallelization of spatial
stochastic reaction-diffusion in STEPS. BMC Neuroscience 2014, 15(Suppl 1):P177.
P55
Accurate approximation to stochastic reaction diffusion on
unstructured meshes in STEPS
Iain Hepburn1,2*, Weiliang Chen1, Erik De Schutter1,2
1
Theoretical Neurobiology, University of Antwerp, 2610 Antwerpen, Belgium;
2
Computational Neuroscience Unit, Okinawa Institute of Science and
Technology Graduate University, Okinawa 904-0495, Japan
E-mail: ihepburn@oist.jp
BMC Neuroscience 2015, 16(Suppl 1):P55
Parallelization is vital in the future of spatial stochastic simulation and an
important component of the Human Brain Project [1]. There are some
unique challenges to be faced for simulators based on unstructured meshes
such as STEPS [2] and finding the most accurate yet scalable solution for
different biological systems is an ongoing area of exploration.
Scalable parallel solutions are known to require an approximation to
diffusion that involves global execution of diffusion events at predetermined
times, an approach that is termed Operator Splitting and was first applied
for regular grids such as by the Gillespie-Multi-Particle method (GMP) [3].
Even serial implementations of GMP show a performance gain compared to
exact solutions, and ~2 orders of magnitude gain has been demonstrated in
a GPU implementation in a simple test system [4]. Therefore, the Operator
Splitting method provides the best known parallel solution. However, it
must be able to perform accurately in comparison to the exact spatial SSA
for a wide range of simulation conditions, and this has not been extensively
investigated before.
We will describe some methods of improving serial Operator Splitting
performance and their effect on accuracy. For example, multinomial
direction selection [5], shows some performance gain in our implementation
compared to uniform random distribution, but depends strongly on the
number of diffusing molecules. This suggests the best approach is to
introduce an adaptive algorithm that can switch between solutions
depending on simulation conditions. Further we find that, to preserve noise
in the system, it is essential to maintain multinomial distribution rather than
apply coarser approximations.
Execution of reaction events must be carefully considered in the Operator
Splitting method and unique solutions are required to avoid introduction
Page 46 of 200
of significant errors. Some previously proposed solutions perform poorly
under sensitive conditions. We will discuss the origin of this error and
demonstrate how these errors can be reduced by maintaining alignment
to the SSA clock. Further, we find convergence of serial Operator Split to
the exact spatial SSA with decreasing communication time-step, which
however comes at a cost to performance.
Our proposed solution draws on many of these ideas, making several
adaptations to simulation conditions, and is tailored for tetrahedral meshes
where probability of leap varies locally. As such, our solution has similarities
to the MPD-RDME method discussed in [6], but with time-steps adaptive to
the upper-limit in the most spatially resolved region, whilst maintaining
alignment to the SSA reaction clock. We will demonstrate accuracy by
application to our routine reaction-diffusion validation tests [2].
References
1. The Human Brain Project. [http://www.humanbrainproject.eu].
2. Hepburn I, Chen W, Wils S, De Schutter E: STEPS: efficient simulation of
stochastic reaction-diffusion models in realistic geometries. BMC Syst Biol
2012, 6:36.
3. Vidal Rodriguez J, Kaandorp JA, Dobrzynski M, Blom JG: Spatial stochastic
modelling of the phosphoenolpyruvate-dependent phosphotransferase
(PTS) pathway in Escherichia coli. Bioinformatics 2006, 22(15):1895-1901.
4. Vigelius M, Lane A, Meyer B: Accelerating reaction-diffusion simulations
with general-purpose graphics processing units. Bioinformatics 2011,
27(2):288-290.
5. Lampoudi S, Gillespie DT, Petzold L: The multinomial simulation algorithm
for discrete stochastic simulation of reaction-diffusion systems. J Chem
Phys 2009, 130:094104.
6. Roberts E, Stone JE, Luthey-Schulten Z: Lattice Microbes: High-
Performance Stochastic Simulation Method for the Reaction-Diffusion
Master Equation. J Comp Chem 2013, 34:245-255.
P56
ATP consumption in molecular signaling of CA1 Hippocampus neurons
Nikon Rasumov*, Erik De Schutter
Okinawa Institute of Science and Technology Graduate University, Onna,
Okinawa 904 0495, Japan
E-mail: nikon@oist.jp
BMC Neuroscience 2015, 16(Suppl 1):P56
The human brain consumes 10 6 times less energy than the currently
fastest super computer [1], while maintaining a comparable performance
in many demanding tasks [2]. This energetic efficiency has been suggested
to result from primitive computations on a molecular level [3]. However,
while the importance of ion channels on energy efficiency has been the
primary focus of research [4,5], most computations occur at the molecular
level prior to the amplification step and prior to the information
transmission through neurons. We calculate the amount of energy
consumed by such computations and compare their structural and
functional properties. As a starting point, we chose 2000 reactions in the
signaling pathways of CA1 hippocampal neurons [6]. As not every reaction
consumes either one or zero ATPs, we undergo a wide literature search to
identify the exact energy consumption of over 60 million of possible
feedback loops. We find that the number of ATPs consumed is related with
size of positive feedback loop. Hence, this study provides the first
systematic and detailed attempt to investigate the energy consumption of
information-storing primitive computations and points towards energy
efficient motifs for synthetic biology.
Acknowledgements: This work was supported by funding from the De
Schutter Unit Okinawa Institute of Science and Technology Graduate
University.
References
1. Niven JE, Laughlin SB: Energy limitation as a selective pressure on the
evolution of sensory systems. Journal of Experimental Biology 2008,
211(11):1792-1804.
2. Ferrucci DA: Introduction to This is Watson. Ibm Journal of Research and
Development 2012, 56(3-4):15.
3. Mead C: Neuromorphic Electronic Systems. Proceedings of the IEEE 1990,
78(10):1629-1636.
4. Sengupta B, Stemmler M, Laughlin SB, Niven JE: Action Potential Energy
Efficiency Varies Among Neuron Types in Vertebrates and Invertebrates.
PLoS Computational Biology 2010, 6(7).
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
5. Alle H, Roth A, Geiger JRP: Energy-Efficient Action Potentials in
Hippocampal Mossy Fibers. Science 2009, 325(5946):1405-1408.
6. Maayan A, Cecchi GA, Wagner J, Rao AR, Iyengar R, Stolovitzky G: Ordered
cyclic motifs contribute to dynamic stability in biological and
engineered networks. Proc Natl Acad Sci U S A 2008, 105(49):19235-19240.
P57
Dynamic model of whole cortex reveals disassortative hub structure in
the intracortical connectome
Matthieu Gilson1*, Ruben Moreno-Bote2,3, Adrian Ponce-Alvarez1, Petra Ritter4,
Gustavo Deco1
1
Center for Brain Cognition, Universitat Pompeu Fabra, 08018, Barcelona,
Spain; 2Parc Sanitari Sant Joan de Du and Universitat de Barcelona,
Esplugues de Llobregat, 08950, Barcelona, Spain; 3CIBERSAM, Esplugues de
Llobregat, 08950 Barcelona, Spain; 4Bernstein Center for Computational
Neuroscience, Berlin, 10117, Germany
E-mail: matthieu.gilson@upf.edu
BMC Neuroscience 2015, 16(Suppl 1):P57
The brain exhibits complex spatio-temporal patterns of activity. In
particular, its baseline neural activity for idle subjects or animals has a
specific structure, also called functional connectivity (FC): cortical areas
experience correlated fluctuations at rest as revealed by imaging
techniques (e.g., fMRI, EEG and MEG). Moreover, these correlation
patterns become more expressed or suppressed depending on the
engaged task. The putative function for this stereotypical background
activity has been actively questioned by the neuroscience community.
A Bayesian hypothesis is that it reflects the variety of possible coactivations
of cortical areas for usual tasks, forming a prior representation of the
environment and interactions wherewith.
The present study relies on our recently developed model [1] in which
noise diffusion is constrained by intracortical white-matter projections to
reproduce FC observed in experimental data (fMRI BOLD signal). In this
model, noise has a functional role and represents the variability of neural
activity during tasks. This model allows us to explore the role of various
parameters in shaping FC, such as the local nonlinear dynamics of cortical
areas. More importantly, we can infer the efficacies of intracortical
projections, also called effective connectivity (EC), from experimental data.
EC plays a crucial role in shaping FC and differs from the structural
connectivity (SC) that is typically obtained using diffusion tensor imaging
(DTI), which estimates the density of white-matter fibers. It is important to
notice that the efficacies of these fibers may differ from their density, due
to the concentration of synapses formed at their end, the type of
neurotransmitters associated and the excitability of target neural
populations. In the end, our model combines anatomical SC and activity
FC to evaluate what drives the neural dynamics, embodied in EC.
Practically, directed EC can be estimated using two matrices of empirical
FC, namely FC0 for correlations with zero time shift and FC for
correlations with a given time shift >0. We apply our method to infer
the intracortical EC from experimental fMRI data that consists of
~20 minutes of recording for 25 idle subjects. The model reproduces
both empirical FC0 and FC matrices with a Pearson correlation
coefficient of 0.65. The recovered EC connectivity is significantly
asymmetric, with a value of 0.3 for an index that scales from 0 for
symmetric matrices to 1 for asymmetric matrices. Moreover, EC exhibits
hubs that have either strong incoming or outgoing weights, but not
both. This disassortative property is expected to exert a strong influence
in shaping the whole network dynamics.
Acknowledgements: MG and GD were supported by the Human Brain
Project. RM-B was supported by the Ramon y Cajal Spanish Award (RYC-
2010-05952), the Marie Curie FP7-PEOPLE-2010-IRG grant (PIRG08-GA-2010-
276795), and the Programa Estatal de Investigacion Cientfica y Tecnica de
Excelencia (grant PSI2013-44811-P). PR acknowledges the support of the
James S. McDonnel Foundation (Brain Network Recovery Group
JSMF22002082), the German Ministry of Education and Research (Bernstein
Focus State Dependencies of Learning 01GQ0971) and the Max-Planck
Society (Minerva Program).
Reference
1. Deco G, Ponce-Alvarez A, Hagmann P, Romani G, Mantini D, et al: How
local excitation-inhibition ratio impacts the whole brain dynamics. J
Neurosci 2014, 34:7886-7898.
Page 47 of 200
P58
The neurodynamical basis of multi-item working memory capacity:
sequential vs simultaneous stimulation paradigms
Marta Balagu1,2*, Laura Dempere-Marco2
1
Moiss Broggi Hospital, Consorci Sanitari Integral, Sant Joan Desp, 08970,
Spain; 2Department of Information and Communication Technologies,
Universitat Pompeu Fabra, Barcelona, 08018, Spain
E-mail: laura.dempere@upf.edu
BMC Neuroscience 2015, 16(Suppl 1):P58
When investigating multi-item WM, and in contrast to single item
experiments, a decision must be made regarding a key aspect of the
stimulation protocol: how the memory set is presented to the subject
simultaneously or sequentially. It is worth noting that most studies
investigating multi-item WM do not address this issue and focus either in
simultaneous stimulation protocols (e.g. [1,2]) or in sequential stimulation
protocols (e.g. [3]) without confronting the two situations. This is
nevertheless an aspect which provides a benchmark to probe and
compare the different theories regarding how resources are allocated
among the different items of a memory set [4,5]. In this study, we explore
a biophysically-realistic attractor model of visual working memory (VWM)
endowed with synaptic facilitation and investigate what are the effects of
varying the dynamics of the facilitation process. We find that: 1) it is
possible to reproduce experimentally observed effects such as the recency
effect in sequential stimulation protocols (i.e. items presented in the final
positions of a sequence are more likely to be retained in WM), and 2) WM
capacity is boosted in both sequential and stimulation protocols when
endowing the attractor network with synaptic facilitation.
Conclusions: In agreement with our previous results [2], synaptic facilitation
boosts the WM capacity limit by effectively increasing the synaptic strengths
just for those pools to which a cue is applied, and then maintaining the
synaptic facilitation by the continuing neuronal firing in only these pools
when the cue is removed. In this study, the time constant F of the synaptic
facilitation process has been found to play an important role in modulating
this effect with large F values leading to larger capacity limits in both
sequential and simultaneous stimulation protocols. However, too large F
values lead to neuronal dynamics which are not compatible with the recency
effect, thus constraining the range of values that F may take.
Acknowledgements: The authors acknowledge funding from the
research project TIN2013-40630-R (Spanish Ministry of Economy and
Competitiveness).
Figure 1(abstract P58) Maintenance of an item in WM memory as a
function of its position within a sequence. The results are derived
from computational simulations (100 blocks of 100 trials) of a delayed
match-to-sample task (same stimulation protocol as in [3] and test item
assimilated to a delayed match-to-sample task) with 9 selective neural
assemblies sequentially stimulated. Maintenance in WM is estimated by
assuming that an item is held in memory when its associated selective
pool shows a mean persistent activity 30 Hz during a period of
500 ms 2 s after the end of the last stimulation. The network parameters
can be found in [2] and F=750 ms in this example
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
References
1. Dempere-Marco L, Melcher DP, Deco G: Effective Visual Working Memory
Capacity: An Emergent Effect from the Neural Dynamics in an Attractor
Network. PLoS One 2012, 7(8):e42719.
2. Rolls ET, Dempere-Marco L, Deco G: Holding Multiple Items in Short Term
Memory: A Neural Mechanism. PLoS One 2013, 8(4):e61078.
3. Amit D, Bernacchia A, Yakovlev V: Multiple-object working memory a
model for behavioral performance. Cereb Cortex 2003, 3:435-443.
4. Bays P, Husain M: Dynamic shifts of limited working memory resources in
human vision. Science 2008, 321:851-854.
5. Zhang W, Luck S: Discrete fixed-resolution representations in visual
working memory. Nature 2008, 453:233-235.
P59
Firing rate response of neocortical neurons in the fluctuation-driven
regime
Yann Zerlaut*, Gilles Ouanounou, Bartosz Telenczuk, Charlotte Deleuze,
Thierry Bal, Alain Destexhe
Unit de Neurosciences, Information et Complexit, CNRS, 91198 Gif sur
Yvette, France
E-mail: zerlaut@unic.cnrs-gif.fr
BMC Neuroscience 2015, 16(Suppl 1):P59
Characterizing the input-output properties of neocortical neurons is of
crucial importance to understand the properties emerging at the network
level. While deriving those input-output relations for artificial neuronal
models has been the object of intense investigations, determining those
properties for real neocortical cells remains both experimentally and
theoretically challenging [1].
Here, we identify four somatic variables that characterize the dynamical
state at the soma in the fluctuation-driven regime and we investigate the
firing rate response of pyramidal neocortical neurons in this four
dimensional space by means of perforated patch recordings in mice
cortex in vitro.
We compare our measurements with different single compartment models
(Integrate and Fire, Exponential Integrate and Fire and Hodgkin-Huxley
models) and we discuss the limitations imposed by those models in the
input-output relationship. We also construct an analytical template for the
firing rate response and we show that it is able to capture the behavior of
both neuronal models and neocortical neurons.
Because of the difficulty to reproduce the measured characteristics with
single compartment models, our results argue in favor of phenomenological
models for the cellular computation, where the complex details of the
biophysical features are absorbed within a flexible higher-level description.
Moreover, the transfer functions obtained from neurons in slices will
enable building mean-field models based on realistic neuronal properties.
Reference
1. La Camera G, Giugliano M, Senn W, Fusi S: The response of cortical
neurons to in vivo-like input current: theory and experiment. Biological
Cybernetics 2008, 99(4-5):279-301.
P60
How neuronal correlations affect the LFP signal?
Bartosz Telenczuk*, Alain Destexhe
Unit de Neurosciences, Information & Complexit, Centre National de la
Recherche Scientifique, 91198 Gif-sur-Yvette, France
E-mail: telenczuk@unic.cnrs-gif.fr
BMC Neuroscience 2015, 16(Suppl 1):P60
Local field potential (LFP) remains a prime source of information about the
activity of the living brain. It allows to monitor the ensemble activity of a
large neuronal population at high temporal resolution. Nevertheless, it is
notoriously hard to analyze and interpret in terms of underlying neuronal
processes. Although the LFP is a population phenomenon, most systematic
efforts to understand it concentrated on single or multiple independent
sources. Over past few years it became clear that the activities of individual
neurons in vivo are strongly correlated. Such correlations could be
introduced internally by direct coupling between the neurons and by shared
inputs or externally by incoming stimuli. To study the effect of these
correlations on the LFP, we introduce a simple and analytically tractable LFP
Page 48 of 200
model. This model consists of a population of correlated Poisson neurons.
Spikes of these neurons contribute a fixed amount to the LFP signal, which
is modeled by convolution of the concatenated spike trains with a temporal
waveform describing the contribution (a LFP kernel). We show that the
cross-correlations between neurons in the population can account for
multiple, seemingly unrelated experimental observations:
1. Multi-unit activity correlates best with high-gamma (60 100 Hz) band of
the LFP [1]. We show that in a population of uncorrelated neurons, the
LFP power is proportional to firing rate, but it scales with its square when
neurons fire correlated spikes. The effect of the correlations on the LFP is
frequency-dependent for realistic values of spike jitter the LFP spectrum
peaks in high-gamma-range frequencies explaining their sensitivity to
modulations of multi-unit activity.
2. Single neuron contribution to LFP estimated from experimental data is
non-causal and spreads over large distances from the neuron [2]. We show
that the estimate of the spike contribution to the LFP, the spike-triggered
LFP average (st-LFP), sums the direct contribution of the trigger neuron
and contributions from all correlated neurons. Since the correlations
extend in time (due to spike jitter) and in space (due to connectivity
patterns), they lead to spatio-temporal smoothing of the direct (causal
and local) contribution.
3. A single spike can initiate a propagating wave in the local neighborhood
of a neuron [3]. We found that the cross-correlogram between pairs of
cortical neurons manifest characteristic widening with distance [2]. We
show that in our model this widening corresponds to increasingly strong
temporal smoothing, which delays the st-LFP peak and produces an
apparent propagation of the spike-evoked field.
4. st-LFP resembles in magnitude and waveform the cross-correlation
between LFP and membrane potential (LFP-Vm) [4]. Both st-LFP and LFP-
Vm are strongly affected by the correlations between neurons. In our
model the st-LFP and LFP-Vm are dissimilar in absence of correlations
and approach each other when correlations gradually increase.
We conclude that the specifics of neuronal correlations can affect the LFP
signals as much as the medium properties or spatial distribution of the
sources. Since the correlations in vivo are often strong and are
modulated by external stimuli and the brain state, such correlations may
be crucial for interpreting the LFP data.
Acknowledgements: Work supported by the CNRS and the European
Commission (Human Brain Project).
References
1. Ray S, Crone NE, Niebur E, Franaszczuk PJ, Hsiao SS: Neural Correlates of
High-amma Oscillations (60-200 Hz) in Macaque Local Field Potentials
and Their Potential Implications in Electrocorticography. J Neurosci 2008,
28:11526-11536.
2. Telenczuk B, Dehghani N, Le Van Quyen M, Cash SS, Halgren E,
Hatsopoulos NG, Destexhe A: On local origins of local field potentials 2015,
(in preparation).
3. Nauhaus I, Busse L, Carandini M, Ringach DL: Stimulus contrast modulates
functional connectivity in visual cortex. Nat Neurosci 2009, 12:70-76.
4. Okun M, Naim A, Lampl I: The subthreshold relation between cortical
local field potential and neuronal firing unveiled by intracellular
recordings in awake rats. J Neurosci 2010, 30:4440-4448.
P61
Effective connectivity analysis explains metastable states of ongoing
activity in cortically embedded systems of coupled synfire chains
Chris Trengove1*, Cees van Leeuwen1, Markus Diesmann2,3,4
1
Perceptual Dynamics Laboratory, University of Leuven, Leuven, Flemish
Brabant, B3000, Belgium; 2Institute of Neuroscience and Medicine (INM-6)
and Institute for Advanced Simulation (IAS-6), Jlich Research Centre and
JARA, Jlich, Germany; 3Department of Psychiatry, Psychotherapy and
Psychosomatics, Medical Faculty, RWTH Aachen University, Aachen, Germany;
4
Department of Physics, Faculty 1, RWTH Aachen University, Aachen,
Germany
E-mail: trengove.c@gmail.com
BMC Neuroscience 2015, 16(Suppl 1):P61
In models of the cortex, synaptic connectivity is often assumed to be
random within broad constraints such as interlayer connection densities.
Alternatively, the connectivity could include richly structured circuitry.
A recent study [1] demonstrated this in a model of local cortex of order
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P61) A The cECG with SCCs and OCs partitioned by their overlaps as nodes. B Chain indices of traversal events versus time (ms),
permuted so chains in each cECG node are adjacent and colored accordingly
1 mm3 in size: a large number of synfire chains (small pools of neurons
sequentially linked by feedforward connections) activated by waves
(sequences of propagating spike packets) were embedded in a recurrent
network of excitatory and inhibitory neurons. The model exhibits stable
global dynamics in the asynchronous irregular regime and stable
propagation of multiple synfire waves. Background noise generated by
waves destabilizes wave propagation, providing a negative feedback
signal limiting their number.
Here we add inter-chain couplings (each chain branches to two successors)
and variability in chain strengths to obtain a recurrent system with a
topography. We study emergent patterns of activity propagation in such a
system. Ongoing endogenous activity due to branching of waves and
regulation of wave activity by noise feedback is typically found. Across
model realizations and runs, ongoing activity manifests diverse steady-
state patterns and transitions. We argue that steady states arise jointly and
consistently with an effective connectivity: strength-dependent chain
traversal probabilities averaged over noise fluctuations. By excluding
chains with sub-threshold traversal probability we derive a family of
effective coupling graphs (ECGs) parameterized by the global activity level.
The distribution of wave activity in steady states is largely confined to the
islands of circulation: strongly connected components (SCCs) and their
associated out-components (OCs) in optimally chosen ECGs. A condensed
ECG (cECG) allows the relationship between activity and effective
connectivity to be visualized (Figure 1).
The system is implemented both as a large-scale network of integrate-and-
fire neurons and as a reduced model with binary-state pools as basic units.
The reduced model exhibits activity patterns very similar to those of the full
model and provides a valuable tool for studying the latter. We propose that
the principle whereby activity patterns arise in concert with dynamically
tuned effective connectivity applies to a broad class of networks with
complex topologies.
Acknowledgements: Partially funded by Helmholtz portfolio theme
SMHB and EU Grants 269921 (BrainScaleS) and 604102 (HBP).
Reference
1. Trengove C, van Leeuwen C, Diesmann M: High capacity embedding of
synfire chains in a cortical network model. J Comput Neurosci 2013,
34(2):185-209.
P62
Deterministic neural networks as sources of uncorrelated noise for
probabilistic computations
Jakob Jordan1*, Tom Tetzlaff1, Mihai Petrovici2, Oliver Breitwieser2,
Ilja Bytschok2, Johannes Bill3, Johannes Schemmel2, Karlheinz Meier2,
Markus Diesmann1
1
Institute for Neuroscience and Medicine (INM-6) and Institute for Advanced
Simulation (IAS-6), Jlich Research Center and JARA, Jlich, Germany;
Page 49 of 200
2
Kirchhoff Institute for Physics, Ruprecht-Karls-University Heidelberg,
Heidelberg, Germany; 3Institute for Theoretical Computer Science, Graz
University of Technology, Graz, Austria
E-mail: j.jordan@fz-juelich.de
BMC Neuroscience 2015, 16(Suppl 1):P62
Neural-network models of brain function often rely on the presence of
noise [1-4]. To date, the interplay of microscopic noise sources and
network function is only poorly understood. In computer simulations and
in neuromorphic hardware [5-7], the number of noise sources (random-
number generators) is limited. In consequence, neurons in large functional
network models have to share noise sources and are therefore correlated.
In general, it is unclear how shared-noise correlations affect the
performance of functional network models. Further, there is so far no
solution to the problem of how a limited number of noise sources can
supply a large number of functional units with uncorrelated noise.
Here, we investigate the performance of neural Boltzmann machines [2-4].
We show that correlations in the background activity are detrimental to
the sampling performance and that the deviations from the target
distribution scale inversely with the number of noise sources. Further, we
show that this problem can be overcome by replacing the finite ensemble
of independent noise sources by a recurrent neural network with the same
number of units. As shown recently, inhibitory feedback, abundant in
biological neural networks, serves as a powerful decorrelation mechanism
[8,9]: Shared-noise correlations are actively suppressed by the network
dynamics. By exploiting this effect, the network performance is
significantly improved. Hence, recurrent neural networks can serve as
natural finite-size noise sources for functional neural networks, both in
biological and in synthetic neuromorphic substrates. Finally we investigate
the impact of sampling network parameters on its ability to faithfully
represent a given well-defined distribution. We show that sampling
networks with sufficiently strong negative feedback can intrinsically
suppress correlations in the background activity, and thereby improve
their performance substantially.
Acknowledgements: Partially supported by the Helmholtz Association
portfolio theme SMHB, the Jlich Aachen Research Alliance (JARA), EU
Grant 269921 (BrainScaleS), The Austrian Science Fund FWF #I753-N23
(PNEUMA), The Manfred Strk Foundation, and EU Grant 604102 (Human
Brain Project, HBP).
References
1. Rolls ET, Deco G: The noisy brain. Oxford University Press 2010.
2. Hinton GE, Sejnowski TJ, Ackley DH: Boltzmann machines: constraint
satisfaction networks that learn. Technical report, Carnegie-Mellon University
1984.
3. Buesing L, Bill J, Nessler B, Maass W: Neural Dynamics as Sampling: A
Model for Stochastic Computation in Recurrent Networks of Spiking
Neurons. PloS CB 2011, 7:e1002211.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
4. Petrovici MA, Bill J, Bytschok I, Schemmel J, Meier K: Stochastic inference
with deterministic spiking neurons 2013, arXiv 1311.3211v1 [q-bio.NC].
5. Schemmel J, Bruederle D, Gruebl A, Hock M, Meier K, Millner S: AWafer-
Scale Neuromorphic Hardware System for Large-Scale Neural Modeling.
Proceedings of the 2010 International Symposium on Circuits and Systems
(ISCAS), IEEE Press 2010, 1947-1950.
6. Bruederle D, Petrovici M, Vogginger B, Ehrlich M, Pfeil T, Millner S, Gruebl A,
Wendt K, Mueller E, Schwartz MO et al: A comprehensive workflow for
general-purpose neural modeling with highly configurable
neuromorphic hardware systems. Biological Cybernetics 2011, 104:263-296.
7. Petrovici MA, Vogginger B, Mueller P, Breitwieser O, Lundqvist M, Muller L,
Ehrlich M, Destexhe A, Lansner A, Schueffny R, et al: Characterization and
Compensation of Network-Level Anomalies in Mixed-Signal
Neuromorphic Modeling Platforms. PLoS ONE 2014, 9(10):e108590.
8. Renart A, De La Rocha J, Bartho P, Hollender L, Parga N, Reyes A, Harris KD:
The asynchronous State in Cortical Circuits. Science 2010, 327:587-590.
9. Tetzlaff T, Helias M, Einevoll G, Diesmann M: Decorrelation of neural-
network activity by inhibitory feedback. PloS CB 2012, 8:e1002596.
P63
Using dynamic time warping for quantifying effects of sinusoidal
oscillation deviations during EEG time series prediction and for finding
interesting EEG and fMRI changes
Dinov Martin
Computational, Cognitive and Clinical Neuroimaging Laboratory, Imperial
College London, W2 0NN, UK
E-mail: m.dinov13@imperial.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P63
Dynamic time warping (DTW) has successfully been used in feature
extraction and analysis of brain data [1,2]. However, DTW and other
dynamic programming (DP) methods are not nearly as widespread in
neuroscientific contexts, as they are in bioinformatics, where they are
central [3]. DP algorithms are extremely powerful and efficient [4], so we
explored further uses of DTW in the analysis of brain data. The first part a)
applies DTW to successfully find locally interesting signal changes in EEG
and functional magnetic resonance imaging (fMRI) data. In the second part
b) we focused on harmonic analysis. Harmonic analysis as conducted via
Fourier Transform (FT) or Wavelet Transform (WT) is extremely widespread
in analyzing brain data, especially electroencephalography (EEG) data. The
fundamental assumption of these methods is that a complex function (the
time series waveform) can be represented as a superposition of sinusoidal
oscillations. While comparing spectra is useful, it is incorrect to assume
that the entirety of the spectral density is due to oscillatory activity proper
(especially sinusoidal) [5]. We wanted to get some idea of the amount and
effects of non-sinusoidal oscillations in EEG data.
In a) we quantify the amount of local signal dissimilarity, which is often
equivalent to where there is interesting sudden activity, and in b) we
quantify how much non-sinusoidal activity there is in spectra computed in
a standard way. For a) we apply the method on a data set from a
simultaneous EEG/fMRI session with 18 subjects that has been used in
other published works before [6] and show that we can easily detect
movement, changes in brain state accompanying large reaction time
changes during a choice reaction time task, and other interesting events in
the EEG and fMRI data sets. The method is applicable to both modalities,
as DTW is very general and can be used with any type of time series. The
method we used works by either running DTW on a selected time interval
(or volume) against sequential intervals (or volumes), or by running it as a
sliding-window for detecting sudden local signal changes in the EEG or
fMRI. For b) we show that the spectral power, as its typically computed,
includes non-sinusoidal activity, when compared to the signal compared
against a series of equal-length (to the signal) sinusoidal waves of varying
frequencies. We also found when the sinusoidal-oscillation deviations
occurred during an EEG time series extrapolation which we performed
using a deep learning Toolbox. The EEG time series prediction gave
another point of view for the oscillation changes. We find that the
prediction breakdown points are related to sudden oscillation changes,
which are presumably time points when other neural oscillator generators
become active and mixed in the signal. We interpret both prediction
breakdowns and momentary oscillatory deviations as being possible
reflections of sudden phase changes in the local or global brain state
Page 50 of 200
related, which may be related to EEG microstate changes. We hope to
bring to attention the flexibility and power of DP to the neuroscientific
community.
Acknowledgements: We thank Dr. Robert Leech for useful discussions
and comments.
References
1. Karamzadeh N, Medvedev A, Azari A, Gandjbakhche A, Najafizadeh L:
Capturing Dynamic Patterns of Task-Based Functional Connectivity with
EEG. Neuroimage 2013, 311-317, February 1.
2. Zheng Y, Liu Q, Chen E, Ge Y, Zhao JL: Time Series Classification Using
Multi-Channels Deep Convolutional Neural Networks. WAIM 2014, LCNS
2014, 8485:298-310.
3. Giegerich R: A systematic approach to dynamic programming in
bioinformatics. Bioinformatics 2000, 16(8):665-677.
4. Bertsekas DP: Dynamic Programming and Optimal Control. Athena
Scientific , 4th 2007, ISBN: 1-886529-26-4.
5. Miller KJ, Sorensen LB, Ojemann JG, Nijs Md: Power-Law Scaling in the
Brain Surface Electric Potential. PLoS Comput Biol 5(12):e1000609,
doi:10.1371/journal.pcbi.1000609.
6. Hellyer PJ, Shanahan M, Scott G, Wise RJS, Sharp DJ, Leech R: The Control
of Global Brain Dynamics: Opposing Actions of Frontoparietal Control
and Default Mode Networks on Attention. The Journal of Neuroscience
2014, 34(2):451-461, 8 January doi: 10.1523/JNEUROSCI.1853-13.2014.
P64
A full rat-scale model of the basal ganglia and thalamocortical network
to reproduce Parkinsonian tremor
Jan Moren*, Jun Igarashi, Junichiro Yoshimoto, Kenji Doya
Neural Computation Unit, Okinawa institute of Science and Technology
Graduate University, Okinawa, Japan
E-mail: jan.moren@gmail.com
BMC Neuroscience 2015, 16(Suppl 1):P64
We aim to investigate the precise mechanisms underlying Parkinsonian
symptoms by large-scale simulation of realistic neural network models
including the basal ganglia (BG), thalamus, the motor cortex and spinal
cord.
Parkinson s disease is a mid- or late-age degenerative disorder of the central
nervous system. The symptoms include akinesia, resting tremor, rigidity and
gait problems. A principal cause of PD is the loss of dopaminergic neurons
in the substantia nigra pars compacta. How the loss of dopamine
transmission causes the symptoms remains unclear.
Although it was previously assumed that the tremor is caused by
enhanced striatum->GPe (external Globus Pallidus) inhibition, recent
experimental data suggests that enhanced GPe->STN (Subthalamic
nucleus) inhibition generates synchronized STN rebound bursts to initiate
tremor [1]. While the BG oscillation is in beta-band (8-15Hz), the final
physical tremor is 4-8Hz, half the BG frequency. The oscillation thus
undergoes sub-harmonic responses downstream of the BG, likely in the
thalamocortical network.
As the oscillation originates through intercellular interactions and depends
on multiple distinct areas, we connect multiple spatially-organized network
models to form an integrated simulation system. We construct separate BG
and thalamocortical models using NEST and connect them with MUSIC.
The separate models enable concurrent development by different
researchers, and the clean separation simplifies the integration work.
The BG model is based on Shouno et al. [2,3]. All subareas are modeled at
the full spatial size and neuron numbers of the rat. The striatum includes
medium-spiny D1 and D2 neurons and fast-spiking interneurons, with D1
neurons projecting directly to the GPi (internal Globus Pallidus) and D2
neurons to the GPe-STN indirect pathway. The total number of neurons
exceeds 3 million. The neuron models are conductance-based with static
or STDP synapses.
The thalamocortical network model consists of four neuron types in the
thalamus and five layers of the primary motor cortex, with neuron
numbers, spatial layout and connections based on experimental data. The
cortical surface size is 3mm2 and consists of ~180K neurons organized in a
stochastic 3-d configuration using Integrate-and-Fire neurons.
We show the BG in a normal and a PD-like state with a ~14Hz oscillation
in the GPi arising from STN-GPe interaction. Inhibitory GPi-thalamus
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
connection may evoke rebound oscillation in the thalamus, which causes
oscillatory output from the cortical pyramidal tract neurons.
References
1. Tachibana Y, Iwamuro H, Kita H, Takada M, Nambu A: Subthalamo-pallidal
interactions underlying parkinsonian neuronal oscillations in the primate
basal ganglia. Eur J Neurosci 2011, 34:1470-1484.
2. Shouno O, Takeuchi J, Tsujino H: A spiking neuron model of the basal
ganglia circuitry that can generate behavioral variability. The Basal
Ganglia IX 2009, 58:, 191-200.
3. Shouno O, Doya K: Local circuit model of the subthalamo-pallidal
network for the generation of parkinsonian oscillations. BMC
Neuroscience 2014, 15(Suppl 1):P168.
P65
Should Hebbian learning be selective for negative excess kurtosis?
Claudius Gros*, Samuel Eckmann, Rodrigo Echeveste
Institute for Theoretical Physics, Goethe University, Frankfurt, 111932 ,
Germany
E-mail: gros07@itp.uni-frankfurt.de
BMC Neuroscience 2015, 16(Suppl 1):P65
Within the Hebbian learning paradigm, synaptic plasticity results in
potentiation whenever pre- and postsynaptic activities are correlated, and
in depression otherwise. This requirement is however not sufficient to
determine the precise functional form for Hebbian learning, and a range of
distinct formulations have been proposed hitherto. They differ, in
particular, in the way runaway synaptic growth is avoided; by either
imposing a hard upper bound for the synaptic strength, overall synaptic
scaling, or additive synaptic decay [1]. Here we propose [2] a multiplicative
Hebbian learning rule which is, at the same time, self-limiting and selective
for negative excess kurtosis (for the case of symmetric input distributions).
Hebbian learning results naturally in a principal component analysis (PCA),
whenever one is present. Alternative formulations of the Hebbian learning
paradigm differ however in other properties. Importantly they may, or may
not, perform an independent component analysis (ICA), whenever one is
feasible. The ICA may be achieved by maximizing (minimizing) the excess
kurtosis, whenever the latter is positive (negative) [3]. Noting that naturally
occurring individual cell lifetime firing rates are, however, characterized by
having both a large kurtosis and a large skewness [4], we investigate in
this paper the effect of the skewness and kurtosis on performing both PCA
and ICA (see Figure 1) with several Hebbian learning rules. A particular
emphasis is placed on the differences between additive and multiplicative
Figure 1(abstract P65) The learning rules ability to perform an
independent component analysis is tested with the non-linear bars
problem. An input set consisting of a random number of horizontal
and vertical bars is fed to the neuron which, after training, becomes
selective to individual bars or pixels, the independent components of
the input set, even when such an input was never presented to the
neuron in isolation
Page 51 of 200
schemes. We find that multiplicative Hebbian plasticity rules select both
for small excess kurtosis and large skewness, allowing them to perform an
ICA, in contrast to additive rules.
Acknowledgements: The support of the German Science Foundation (DFG)
and the German Academic Exchange Service (DAAD) are acknowledged.
References
1. Abbott LF, Nelson SB: Synaptic plasticity: taming the beast. Nat Neurosci
2000, 3:1178-1183.
2. Echeveste R, Gros C: Generating Functionals for Computational
Intelligence: The Fisher Information as an Objective Function for Self-
Limiting Hebbian Learning Rules. Front Robot AI 2014, 1:1.
3. Hyvrinen A, Oja E: Independent component analysis: algorithms and
applications. Neural Netw 2000, 13:411-430.
4. Willmore B, Tolhurst DJ: Characterizing the sparseness of neural codes.
Network-Comp Neural 2001, 12:255-270.
P66
Decoding position from multiunit activity using a marked point process
filter
Xinyi Deng1*, Daniel F Liu2, Kenneth Kay2, Loren M Frank3, Uri T Eden1
1
Department of Mathematics and Statistics, Boston University, Boston, MA,
USA; 2UC BerkeleyUCSF Graduate Program in Bioengineering, UCSF, San
Francisco, CA, USA; 3Department of Physiology, UCSF, San Francisco, CA, USA
E-mail: xinyi@math.bu.edu
BMC Neuroscience 2015, 16(Suppl 1):P66
Traditionally, experiments designed to study the role of specific spike
patterns in learning and memory tasks take one of two forms, 1)
observational studies that characterize statistical properties of neural activity
during such tasks or 2) interventional studies that broadly alter neural
activities over an entire neural population or brain region. This work is part
of a larger project to allow investigators to manipulate neural populations in
a content-specific way, altering spiking activity related to certain learning
and memory patterns while leaving activity related to other patterns intact.
One fundamental challenge of this work is to decode the information
content of specific spike sequences in real-time. Previously, we have used
point process theory to develop efficient decoding algorithms based on
spike train observations. However these algorithms assume the spike trains
have been accurately sorted ahead of time, which is not possible for real-
time decoding. Here we present a new point process decoding algorithm
that does not require multiunit signals to be sorted. We use the theory of
marked point processes to characterize the relationship between the coding
properties of multiunit activity and features of the spike waveforms [1-3].
Using Bayes rule, we compute the posterior distribution of a signal to
decode given multiunit activity from a neural population.
We first characterize the spiking activity of a neural population using the
conditional intensity function for marked point processes. We then
construct point process filters to iteratively calculate the full posterior
density of a signal. We illustrate our approach with a simulation study as
well as with experimental data recorded in the hippocampus of a rat
performing a spatial memory task. Our decoding framework is used to
reconstruct the animals position from unsorted multiunit spiking activity.
We then compare the quality of fit of our decoding framework to that of
a traditional spike-sorting and decoding framework. Our analyses show
that the proposed decoding algorithm performs as well as or better than
algorithms based on sorted single-unit activity. These results provide a
mechanism for content-specific manipulations of population activity in
hippocampus.
Acknowledgements: This research was supported in part by a grant
from the Simons Foundation (SCGB 320135 to L.M.F. and SCGB 337036 to
U.T.E.), NSF grant IIS-0643995 and NINDS grant R01 NS073118 to U.T.E.,
NIH grant R01 MH0901188 to L.M.F., and NSF GRFP grant No. 1144247 to
D.F.L
References
1. Chen Z, Kloosterman F, Layton S, Wilson MA: Transductive Neural
Decoding for Unsorted Neuronal Spikes of Rat Hippocampus. Proceedings
of the 34th Annual International Conference of the IEEE EMBS 2012.
2. Cox DR, Isham V: Point Processes London: Chapman and Hall 1980.
3. Kloosterman F, Layton S, Chen Z, Wilson MA: Bayesian decoding using
unsorted spikes in the rat hipposcampus. J Neurophysiol 111:217-227.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
P67
Hybrid scheme for modeling local field potentials from point-neuron
networks
Espen Hagen1,2*, David Dahmen1, Maria Stavrinou2, Henrik Lindn3,4,
Tom Tetzlaff1, Sacha van Albada1, Sonja Grn1,5, Markus Diesmann1,6,7,
Gaute T Einevoll2,8
1 1. Lindn H, Tetzlaff T, Potjans TC, Pettersen KH, Gruen S, Diesmann M,
Inst. of Neuroscience and Medicine (INM-6) and Inst. for Advanced
Simulation (IAS-6), Jlich Research Center and JARA, Jlich, 52425, Germany;
2
Dept. of Mathematical Sciences and Technology, Norwegian University of
Life Sciences, Aas, 1432, Norway; 3Dept. of Neuroscience and Pharmacology,
University of Copenhagen, Copenhagen, 2200, Denmark; 4Dept. of
Computational Biology, Royal Institute of Technology (KTH), Stockholm,
10044, Sweden; 5Dept. of Biology, Theoretical Systems Neurobiology, RWTH
Aachen University, Aachen, 52074, Germany; 6Dept. of Psychiatry,
Psychotherapy and Psychosomatics, Medical Faculty, RWTH Aachen
University, Aachen, 52074, Germany; 7Dept. of Physics, Faculty 1, RWTH
Aachen University, Aachen, 52074, Germany; 8Dept. of Physics, University of
Oslo, Oslo, 0316, Norway
E-mail: e.hagen@fz-juelich.de
BMC Neuroscience 2015, 16(Suppl 1):P67
Measurement of the local field potential (LFP) has become routine for
assessment of neuronal activity in neuroscientific and clinical applications,
but its interpretation remains nontrivial. Understanding the LFP requires
accounting for both anatomical and electrophysiological features of neurons
near the recording electrode as well as the entire large-scale neuronal
circuitry generating synaptic input to these cells. The direct simulation of
LFPs in biophysically detailed network models is computationally daunting.
Here, we instead propose a hybrid modeling scheme combining the
efficiency of simplified point-neuron network models (Fig. 1A) with the
biophysical principles underlying LFP generation by multicompartment
neurons [1] (Fig 1C). We apply this scheme to a model representing a full-
scale cortical network under about 1 square millimeter surface of cat
primary visual cortex [2] (Fig. 1A,B) with layer-specific connectivity [3] to
predict laminar LFPs (Fig. 1D) for different network states, assess the relative
contribution of local neuron populations to the LFP, investigate the role of
input correlations and neuron density, and validate linear LFP predictions
Figure 1(abstract P67) Overview of the hybrid scheme for modeling LFP generated by a cortical network model. A Sketch of point-neuron
network model [1]. B Spikes of point neurons in the network for spontaneous and evoked activity. C Populations of multi-compartment neurons acting
as LFP signal generators. D Depth-resolved LFP predicted using the hybrid scheme
Page 52 of 200
based on population firing rates. The hybrid scheme is accompanied by our
open-source software, hybridLFPy (github.com/espenhgn/hybridLFPy).
Acknowledgements: This research was partially funded by EU Grant
604102 (HBP), EU Grant 269921 (BrainScaleS), the Helmholtz portfolio
theme SMHB, the Juelich Aachen Research Alliance (JARA), and the
Research Council of Norway (NFR, through ISP, NOTUR -NN4661K).
References
Einevoll GT: Modeling the spatial reach of the LFP. Neuron 2011,
72:859-872.
2. Potjans TC, Diesmann M: The Cell-Type Specific Cortical Microcircuit:
Relating Structure and Activity in a Full-Scale Spiking Network Model.
Cerebral Cortex 2014, 24(3):785-806.
3. Binzegger T, Douglas RJ, Martin KA: A quantitative map of the circuit of
cat primary visual cortex. J Neurosci 2004, 24(39):8441-8453.
P68
Can ionic diffusion have an effect on extracellular potentials?
Geir Halnes1*, Tuomo Mki-Marttunen2, Klas H Pettersen3, Daniel Keller4,
Ole A Andreassen2, Gaute T Einevoll1,5
1
Dept. of Mathematical Sciences and Technology, Norwegian University of
Life Sciences, s, Norway; 2NORMENT, Institute of Clinical Medicine,
University of Oslo, Oslo, Norway; 3Centre for Molecular Medicine Norway,
University of Oslo, Oslo, Norway; 4The Blue Brain Project, EPFL, Lausanne,
Switzerland; 5Department of Physics, University of Oslo, Oslo, Norway
E-mail: geir.halnes@nmbu.no
BMC Neuroscience 2015, 16(Suppl 1):P68
Several pathological conditions, such as hypoxia, anoxia, ischemia and
spreading depression are associated with ion concentration changes in the
extracellular space (ECS) [1]. Also during non-pathological conditions, endured
periods of intense neural signaling may cause local ion concentration changes
in the millimolar range. Changes in ion concentrations are often accompanied
by a slow negative potential shift measured in the ECS [1-3], the origin of
which is not properly understood. In computational neuroscience, it is
common to use the simplifying assumption that diffusive currents (along
concentration gradients) are negligible compared to Ohmic currents (along
electrical fields). This is, for example, an underlying assumption in the cable
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
equation used in most multi-compartmental neural models. It is also an
underlying assumption in standard current source density (CSD) theory, which
predicts neural current sources from recordings of extracellular potentials.
However, theoretical studies have identified scenarios where large, but
biologically realistic, ion concentration gradients may induce diffusive currents
that are comparable in magnitude to Ohmic currents [4,5].
In the current work, we have explored possible effects that diffusive currents
may have on the extracellular potential in a cortical column. To do this, we
developed a hybrid simulation framework: First, we used the simulator
NEURON (and an available multicompartmental neural model with realistic
morphology [6]) to simulate the transmembrane ionic output from a small
population of cortical neurons driven by realistic synaptic input. Next, we
used a separate scheme to predict variations in extracellular potentials and
ion concentrations. The scheme was based on the Nernst-Planck equations
for electrodiffusion and the constraint of electroneutrality, and represents a
novel, efficient and generally applicable method for simulating extracellular
dynamics surrounding multicompartmental neural models or networks of
such (e.g., the Blue Brain simulator).
Our key findings were: (i) Morphology-dependent distributions of
transmembrane current sources/sinks induced sustained extracellular
potentials. Spatial variations of the ECS potential were of the order of a
few mV across the depth of the cortical columns, which is similar to
experimentally observed sustained potential profiles [2,3]. (ii) Long periods
of neuronal output (simulations were run for 80 s) could change local ECS
ion concentrations by several millimolars. (iii) For large, but realistic,
concentration gradients in the ECS, diffusive currents were of the same
magnitude as Ohmic currents. (iv) Diffusive currents could have a
significant impact on the sustained extracellular potentials, and could be a
possible explanation to the observed relationship between ionic shifts and
voltage drops in the ECS [1-3]. (v) Variation in concentrations and diffusive
currents took place at the time scale of several seconds, and had little
impact on the frequencies that are of predominant interest in standard
CSD-analysis.
References
1. Sykov E, Nicholson C: Diffusion in Brain Extracellular Space. Physiol Rev
2008, 88:1277-1340.
2. Cordingley G, Somjen G: The clearing of excess potassium from
extracellular space in spinal cord and cerebral cortex. Brain Research
1978, 151:291-306.
3. Dietzel I, Heinemann U, Lux H: Relations between slow extracellular
potential changes, glial potassium buffering, and electrolyte and cellular
volume changes during neuronal hyperactivity in cat. Glia 1989, 2:25-44.
4. Qian N, Sejnowski T: An electro-diffusion model for computing
membrane potentials and ionic concentrations in branching dendrites,
spines and axons. Biological Cybernetics 1989, 15:1-15.
5. Halnes G, stby I, Pettersen KH, Omholt SW, Einevoll GT: Electrodiffusive
model for astrocytic and neuronal ion concentration dynamics. PLoS
Computational Biology 2013, 9(12):e1003386.
6. Hay E, Hill S, Schrmann F, Markram H, Segev I: Models of neocortical
layer 5b pyramidal cells capturing a wide range of dendritic and
perisomatic active properties. PLoS Computational Biology 2011, 7(7):
e1002107.
P69
Attentional spreading over feature attributes and feature dimensions:
Distributed top-down modulation or joint neural coding?
Lisa Bohnenkamp1,2*, Detlef Wegener2, Udo Ernst1
1
Computational Neuroscience, University of Bremen, Bremen, 28359,
Germany; 2Theoretical Neurobiology, University of Bremen, Bremen, 28359,
Germany
E-mail: lisa.bohnenkamp@uni-bremen.de
BMC Neuroscience 2015, 16(Suppl 1):P69
Attention is an important prerequisite for visual information processing.
It allows the brain to focus on particular aspects of a visual scene, and
enables or enhances perception of complex shapes and objects. However,
attentional selection is an intricate computational problem, since the target
of attentional deployment is often given by an arbitrary combination of
features such as the location of an object having a conjunction of particular
properties (features), and it is subjected to various neural constraints such as
a given anatomical connectivity and broad tuning of neurons. How the
Page 53 of 200
brain solves this problem to boost the processing of combinations of
different features that are represented across multiple neural areas is largely
unknown.
Recently, progress in understanding the properties of joint attentional
selection was made in a psychophysical study investigating attentional
spreading within and across objects [1]: Subjects were asked to report
color and speed changes on one of two overlapping random dot
patterns. Only one of the features was unique for each object, while the
other was shared by both. Reaction times (RTs) recorded under different
cueing conditions demonstrated the co-selection of unattended features,
with attention spreading from the attended feature attribute in a
particular feature dimension to other feature attributes and other feature
dimensions. Importantly, this processing benefit was not restricted to the
task-relevant object but extended to the unattended object.
It is an open question how these observations can be modeled and
understood in a coherent framework. In our contribution, we propose two
structurally simple models, implementing two complementary neural
mechanisms: The first model assumes that the specificity of anatomical
connections providing attentional feedback from higher cortical areas is
constrained. Consequently, top-down modulation of neural activity in
lower visual areas is broadly tuned, targeting all cells representing the
cued feature dimension(s) with only a weak preference for the cued
feature attribute(s). As a result, attention spreads to uncued feature
attributes and to jointly cued feature dimensions. In the second model we
assume joint coding of stimulus features such that neurons having a
preference for a particular motion direction also have a preference for a
specific color. We also assume that neurons are not always perfectly tuned
to their preferred feature(s), thus providing a (small) response to non-
preferred feature values as well. Attentional feedback in this model
precisely targets cells tuned for the cued feature attribute, or the
combination of attributes. In this approach, attentional spreading to
uncued feature dimensions is mediated by joint tuning, whereas spreading
to uncued feature attributes in the same feature dimension is mediated by
imperfect tuning.
It turns out that both approaches, with appropriately chosen parameters,
can qualitatively explain the differences in RTs between the stimulation
conditions (Fig. 1). The next step is to develop a neurophysiologically
plausible model that allows for explicit predictions for electrophysiological
experiments to critically test the proposed mechanisms.
Acknowledgements: This work was supported by the BMBF (Bernstein
Award Udo Ernst, grant no. 01GQ1106).
Reference
1. Wegener D, Galashan FO, Aurich MK, and Kreiter AK: Attentional spreading
to task-irrelevant object features: Selection is object-based but
processing modulation is feature-based. Frontiers Hum. Neurosci 2014,
8:414, doi: 10.3389/fnhum.2014.00414.
P70
Probing information routing mechanisms by precisely-timed electrical
stimulation pulses: a modelling study
Dmitriy Lisitsyn*, Daniel Harnack, Udo Ernst
Institute for Theoretical Physics, Computational Neuroscience Lab, University
of Bremen, Bremen, D-28159, Germany
E-mail: dmitriy@neuro.uni-bremen.de
BMC Neuroscience 2015, 16(Suppl 1):P70
In dependence on the behavioral context, sensory signals are selectively
routed throughout the brain. Particularly, in area V4 of the visual cortex,
neurons with multiple stimuli in their receptive fields respond as if just the
attended stimulus was present [1]. The mechanisms for selective
information processing are still debated, but gamma-band synchronization
between brain areas such as V1 and V4 emerged as a promising candidate,
proposing that information is routed between populations with a
synchronized in-phase relationship and blocked when gamma oscillations
are in an anti-phase or random relationship [2] (communication-through-
coherence, CTC).
The causality of this hypothesis has proven hard to establish, but cortical
stimulation gives us the tools to investigate it. By utilizing controlled
perturbations, we could shift neural populations into desired phase
relationships and quantify the effect on visual information gating. As a
precursor to animal experiments, we investigate this paradigm by using
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 54 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

Figure 1(abstract P69) Comparison of Median RT. (A) Experimental data of Exp. 1 from Ref. [1]. (B) Data from first model. (C) Data from second model.
Different cuing conditions on x-axis: C: correct, W: wrong, F: feature, O: object. Left column shows RT to a speed change, right column to a color change

neural models to predict the effect of stimulation on gamma oscillations,
studying the feasibility of different synchronization scenarios and their
robustness against noise.
Without stimulation, the troughs of the unpulsed V4 LFP (in grey) coincide
with the peaks of the V1 population B (in red), hence suppressing the
incoming information from B while attending to the signal carried by A.
However, a pulse applied at the proper phase of the V4 LFP (marked in
green), results in a phase shift of the LFP (in black). The troughs of the new
LFP now coincide with V1 population A, thus effectively changing which
input signal is gated/attended.
First, we explored the effect of precisely-timed electric pulses on different
oscillation mechanisms proposed for generating g-rhythms, specifically ING
and PING [3], for which we obtained phase-response curves (PRC) under
multiple stimulation regimes. Using these oscillation mechanisms, we
constructed a simple network representing two V1 populations A and B,
and one V4 population. When simultaneously driven by two input stimuli,
the two V1 populations would oscillate in anti-phase, while the V4
population would synchronize with V1 populations A and B in an alternate
fashion (bistable dynamics). Based on the computed PRCs, a stimulation
paradigm was developed which monitored the relative phase relations of
the populations online, allowing us to pulse V4 at the right time to
successfully control which V1 population the V4 population would
synchronize with, thus enhancing the information transfer between these
two modules (see Figure 1).
The success of stimulation strongly depends on the noise level and the
resulting reliability of Gamma oscillations. At a critical point, shifting
phases gradually is no longer feasible and a stronger phase-reset pulse has
to be used instead. In addition, a careful monitoring of on-going activity is
essential since phase control is most efficient in periods where the
observed Gamma amplitude is high.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P70) LFP and spike data of the model. Without stimulation, the troughs of the unpulsed V4 LFP (in grey) coincide with the peaks of
the V1 population B (in red), hence suppressing the incoming information from B while attending to the signal carried by A. However, a pulse applied
at the proper phase of the V4 LFP (marked in green), results in a phase shift of the LFP (in black). The troughs of the new LFP now coincide with V1
population A, thus effectively changing which input signal is gated/attended
Acknowledgements: This work was supported by BMBF (Bernstein
Award Udo Ernst, grant no. 01GQ1106).
References
1. Moran J, Desimone R: Selective attention gates visual processing in the
extrastriate cortex. Science 1985, 229(4715):782-784.
2. Fries P: A mechanism for cognitive dynamics: neuronal communication
through neuronal coherence. Trends in cognitive sciences 2005,
9(10):474-480.
3. Tiesinga P, Sejnowski TJ: Cortical enlightenment: are attentional gamma
oscillations driven by ING or PING? Neuron 2009, 63(6).
P71
Area summation is related to efficient neural representation
Fariba Sharifian1,2,3,4*, Hanna Heikkinen1,2, Ricardo Vigrio5, Simo Vanni3,4
1
Brain Research Unit, Department of Neuroscience and Biomedical
Engineering, Aalto University, Espoo, Finland; 2AMI Centre, Aalto
Neuroimaging, Aalto University, Espoo, Finland; 3Clinical Neurosciences,
Neurology, University of Helsinki, Helsinki, Finland; 4Helsinki University
Hospital, Helsinki, Finland; 5Department of Information and Computer
Science, Aalto University School of Science, Aalto University, Espoo, Finland
E-mail: fariba@neuro.hut.fi
BMC Neuroscience 2015, 16(Suppl 1):P71
In the primary visual cortex (V1), increasing stimulus size first increases and
then decreases the neural firing rate before reaching an asymptote [1,2].
This typical response curve, area summation function (ASF), defines
canonical contextual modulation in typical V1 neuron. Previous studies
suggest that contextual modulation associates with efficiency of the neural
network [3,4]. Here, we studied the relationship between ASF and efficiency
of system level neural activation patterns.
Cavanaugh et al. [1] provided quantitative data from macaque cortex, as
well as a mathematical description for the ASF. Therefore, we created a
theoretical expected cortical response for an arbitrary gray-level image
input stimulus based on ASF. Next, we used a biophysically meaningful
(biomimetic) network of exponential integrate-and-fire neurons to
stimulate V1 response (Heikkinen, Sharifian, Vigario and Vanni,
unpublished observations). As an outcome, we compared the distance
between the biomimetic simulator outputs to the expected ASF function
(Distance to Area Summation, DAS). Finally, we studied entropy per spike,
energy consumption and neural population sparseness as a function of the
DAS for 20 natural grayscale images. In the biomimetic simulation we left
V1-extrastriate and V1 excitatory-inhibitory connection strengths as free
parameters in the simulations. A square search with 626 combinations of
values was run for all the 20 natural stimuli resulting in 20 * 626 = 12520
simulated response patterns.
Page 55 of 200
Our results show that: i) there is a clear association between the entropy
per spike and DAS (distance correlation = 0.66, Fig 1.a). ii) A negative link
exists between the level of energy consumption (linearly associated with
spike frequency) in modeled output patterns and DAS (r = 0.62, p < 0.001,
Fig 1.b). iii) Population sparseness of the modeled output patterns is
positively related to the DAS (r = 0.60, p < 0.001, Fig 1.c).
In conclusion, the results of our study suggest that there is a link between
area summation function and efficient information coding in the cortical
visual system.
References
1. Cavanaugh JR, Bair W, Movshon JA: Nature and interaction of signals from
the receptive field center and surround in macaque V1 neurons. Journal
of Neurophysiology 2002, 88(5):2530-2546.
2. Angelucci A, Levitt JB, Walton EJ, Hupe JM, Bullier J, Lund JS: Circuits for
local and global signal integration in primary visual cortex. Journal of
Neuroscience 2002, 22(19):8633-8646.
3. Sharifian F, Nurminen L, Vanni S: Visual interactions conform to pattern
decorrelation in multiple cortical areas. Plos One 2013, 8(7):e68046.
4. Vanni S: Local model for contextual modulation in the cerebral cortex.
Neural Networks 2012, 25:30.
P72
Enhanced novelty detection in auditory scenes through adaptation of
inhibition
Bertrand Fontaine
Laboratory of Auditory Neurophysiology, Faculty of Medicine, University of
Leuven, Leuven, Belgium
E-mail: bertrand.fontaine@med.kuleuven.be
BMC Neuroscience 2015, 16(Suppl 1):P72
Animals and humans must behave efficiently in acoustically rich
environments, where few sound sources are of interest out of many others
that are considered background noise. How does the auditory system
implement such robust processing is a topic of intense research. In this
work I am interested in the neural mechanisms underlying detection of
novel objects in auditory scenes, such as a novel sound source at a given
location.
It has been shown psycho-physically that the perception of a sound
depends on the acoustical context in which it occurs. A basic example of
such effect is called enhancement [1]. A harmonic complex (i.e. a series of
tones at different frequencies) presented in isolation will be perceived as a
single auditory object. If one tone is deleted from the complex (the
conditioner) and then reintroduced (test), it will perceptually pop out as
a separate sound object. To maximally induce enhancement in humans,
the conditioner must have an energy notch in its spectrum with a
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P71)
preferred width of 0.6 octave. This effect originates in the central nervous
system; neural correlate thereof were found in the cochlear nucleus, two
synapses away from the cochlea [2]. The standard model explaining the
emergence of auditory enhancement is based on the adaptation of wide-
band inhibition inputs [1,2]: the increase in output rate due to a preceding
conditioner comes from a decrease in activity of the wide-band inhibitory
inputs, i.e. inhibition have adapted during the conditioner while narrow
band excitation has not. In this study I seek to test functions of
enhancement effects in realistic listening situations. In particular, I
hypothesize that the enhancement of speech signals by realistic
environments is a neural implementation of a novelty detector.
I implemented on the Brian simulator and its auditory package [3] a model
of the auditory periphery followed by a network of adapting spiking
neurons [4] comprising feed forward excitation and inhibition. This network
was first tuned to reproduce the basic enhancement effects reported
physiologically. In order to mimic realistic listening conditions, sounds are
filtered by realistic environmental processes prior to being fed to the model.
In particular Im interested in the filtering induced in sounds arriving at the
ear by the listeners body and by the surrounding room. The head and the
shoulder create reflections of the direct sound, resulting in notches in the
high frequency sound spectrum, which depend on the sound source
elevation [5]. Similarly, reflections from the ground and the walls of a room
will induce notches in the low-frequency spectrum, which depend on the
sound source azimuth and distance [6]. As spectral notches are critical in
inducing auditory enhancement I show that if a novel sound is preceded by
other sounds located at different positions, the detection of this novel
sound will be enhanced.
The present functional study of auditory enhancement suggests that such
contextual effects could be critical for animal survival and for hearing in
realistic conditions. The results show how simple neural mechanisms
modulate a sensory pathway so that it adapts to its surrounding
environment and suggest that bias in perception could have ecological
advantages. The model could also inspire the design of speech processing
devices in real environments.
Acknowledgements: Work supported by a European Community Marie
Curie fellowship (PIOF-GA-2011-300753).
References
1. Carcagno S, Semal C, Demany L: Enhancement of increments in spectral
amplitude: further evidence for a mechanism based on central
adaptation. Adv Exp Med Biol 2013, 787:175-182.
2. Fontaine B, Franken T, Joris PX: Neural correlates of context-dependent
enhancement in the Dorsal Cochlear Nucleus. Assoc Res Otolaryngol 2015,
38:84, Abs.
3. Fontaine B, Goodman DFM, Benichoux V, Brette R: Brian hears: online
auditory processing using vectorization over channels. Front. Neuroinform
2011, 5:9, doi: 10.3389/fninf.2011.00009.
4. Fontaine B, Pea JL, Brette R: Spike-threshold adaptation predicted by
membrane potential dynamics in vivo. PLoS Comput Biol 2014, 10(4):
e1003560.
Page 56 of 200
5. Blauert J: Spatial hearing: the psychophysics of human sound
localization. MIT press 1997.
6. Gourvitch B, Brette R: The impact of early reflections on binaural cues.
J Acoust Soc Am 2012, 132:9-27.
P73
Mechanisms of cortical high-gamma activity (60-200 Hz) investigated
with computational modeling
Piotr Suffczynski1*, Nathan E Crone2, Piotr J Franaszczuk3
1
Department of Biomedical Physics, University of Warsaw, Warsaw, 02-093,
Poland; 2Department of Neurology, The Johns Hopkins University School Of
Medicine, Baltimore, MD 21287, USA; 3Human Research & Engineering
Directorate, U.S. Army Research Laboratory, APG, Aberdeen, MD 21005-5425, USA
E-mail: Piotr.Suffczynski@fuw.edu.pl
BMC Neuroscience 2015, 16(Suppl 1):P73
High-gamma activity (HGA) at frequencies 60-200 Hz have been observed
during task-related cortical activation in humans [1] and in animals [2], and
have been used to map normal brain function and to decode commands
in brain-computer interfaces. To understand the role that HGA plays in
both normal and pathological brain states, deeper insights into its
generating mechanisms are essential. Because the neural populations
recorded by LFPs and EEG cannot be comprehensively recorded at scales
that are likely to be relevant, we used a biologically based computational
model of a cortical network to investigate the mechanisms generating
HGA. The computational model included excitatory pyramidal regular-
spiking and inhibitory fast-spiking neurons described by Hodgkin - Huxley
dynamics. We compared activity generated by this model with HGA that
was observed in LFP recorded in monkey somatosensory cortex during
vibrotactile stimulation. Increase of firing rate and broadband HGA
responses in LFP signals generated by the model were in agreement with
experimental results (see Figure 1).
Conclusions: The HGA appear to be mediated mostly by an excited
population of inhibitory fast-spiking interneurons firing at high-gamma
frequencies and pacing excitatory regular-spiking pyramidal cells, which fire
at lower rates but in phase with the population rhythm. HGA reflects local
cortical activation under normal conditions and as such is a good candidate
for mapping cortical areas engaged by a specific task. The mechanisms of
HGA, in this model of local cortical circuits, appear to be similar to those
proposed for hippocampal ripples generated by subset of interneurons that
regulate discharge of principal cells.
References
1. Crone NE, Sinai AS: A Korzeniewska High-frequency gamma oscillations
and human brain mapping with electrocorticography. Prog Brain Res
2006, 159:275-295.
2. Ray S, Hsiao SS, Crone NE, Franaszczuk PJ, Niebur E: Effect of stimulus
intensity on the spike-local field potential relationship in the secondary
somatosensory cortex. J Neurosci 2008, 28:7334-7343.
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 57 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

Figure 1(abstract P73) Comparison of high-gamma observed in vivo (AB) and simulated in the model (CD) during sensory stimulation for
different stimulus amplitudes denoted G1, G2, G5 and G10. AC: average firing rate of neurons. BD: time - frequency maps of LFP signals
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
P74
A reservoir network model for sensory-guided probabilistic decision
making
Tomoki Kurikawa1*, Takashi Handa2, Tomoki Fukai1
1
Brain Science Institute, RIKEN, Wako, Saitama, 351-0198, Japan; 2Research
Center Caesar, Bonn, 53175, Germany
E-mail: tomoki.kurikawa@riknen.jp
BMC Neuroscience 2015, 16(Suppl 1):P74
Animals are often required to adequately respond to novel stimuli on the
basis of their previous sensory experiences. To investigate the neural
mechanism underlying this sensory-guided decision making, we
conducted multiunit recordings from the rats performing a two-alternative
choice task. The rats were trained to make a LEFT or a RIGHT choice in
response to two auditory cues, and then their behavior was tested for
these familiar cues and other novel cues. Their choice probabilities
generally varied in a graded manner such that the probability of choosing
an option changed gradually according to frequency differences between
familiar and novel cues. However, we also observed large variability in
choice behavior across rats: choice probabilities for novel cues were near
the chance level in some rats, while it systematically changed with tone
frequency in other rats. We elucidated the mechanism underlying such
decision behavior and the possible origin of individual behavioral
differences in a neural network model (Fig. 1A). Our model is viewed as a
reservoir network [1] and learns to associate two familiar cues with two
alternative choices through reinforcement learning with eligibility trace [2].
Our model successfully replicated the gradual choice behavior observed
in our experiment. We further analyzed how neural dynamics in the
reservoir network determines the choice probability for familiar and novel
cues. We found that a familiar stimulus sequentially activates a relatively
small portion of reservoir neurons, and reinforcement learning trained
output connections from these neurons such that only an adequate
output neuron is activated by the neural trajectory evoked in the
reservoir. We further revealed that choice responses to novel cues
become graded due to trial-by-trial overlaps between the familiar
trajectories and novel-cue-evoked trajectories.
Interestingly, our model also exhibited similar variability in choice
responses to that observed in individual rats. We found that if input
neurons are highly sensitive to external stimuli, that is, if the width of
their frequency tuning curves is broad, the model network likely shows
gradual choice behavior for novel stimuli. In contrast, the model with
more sensitive input neurons (with broader tuning curves) tends to
generate near-random choice behavior, displaying a flat dependence of
choice probability on novel stimuli. We compared choice behavior
between the models and the rats by introducing quantitative measures
Figure 1(abstract P74)
Page 58 of 200
and found that the behavioral tendency of our model is consistent with
that of the rats (Fig. 1B).
These results may suggest that some individual differences in decision
making behavior emerge from neural population dynamics rather than
differences in higher-level behavioral strategies.
Acknowledgements: This work was partially supported by KAKEN-HI No.
255413, Grants-in-Aid for Scientific Research (no. 22115013) from MEXT.
References
1. Maass W, Natschlger T, Markram H: Real-Time Computing Without Stable
States: A New Framework for Neural Computation Based on
Perturbations. Neural Comput 2002, 14:2531-2560.
2. Izhikevich EM: Solving the distal reward problem through linkage of
STDP and dopamine signaling. Cereb Cortex 2007, 17:2443-52.
P75
Stimulus induced resonance in a neural mass model driven with a
temporally correlated noise
Maciej Jedynak1,2*, Antonio J Pons1, Jordi Garcia-Ojalvo2
1
Departament de Fsica i Enginyeria Nuclear, Universitat Politcnica de
Catalunya, Barcelona, Spain; 2Department of Experimental and Health
Sciences, Universitat Pompeu Fabra, Parc de Recerca Biomdica de
Barcelona, Barcelona, Spain
E-mail: maciej.jedynak@upf.edu
BMC Neuroscience 2015, 16(Suppl 1):P75
Power spectra of experimental recordings such as EEG and LFP exhibit a
broadband nature characterized by a stable 1/f b background with
embedded peaks. These peaks are associated with specific brain states
and stimuli, and depending on conditions appear at different frequencies.
Interactions between brain rhythms from different frequency bands are
assumed to play an important role in brain functioning. Here we study
theoretically a particular cross-frequency interaction that has been
observed experimentally [1]. That experiment examined the effect of slow
rocking on sleep in human subjects, and showed that slow oscillatory
stimulus related to rocking enhanced power in EEG measurements in low
(0.6-4 Hz) and alpha frequency bands, thus revealing a cross-frequency
transfer of power.
We studied this effect by means of a neural mass description based on a
model developed by Jansen and Rit [2]. This model typically yields a
power spectrum with a narrow peak that corresponds to its intrinsic
frequency. In our case, however, a realistic 1/f b spectrum with an
embedded alpha peak is observed. This results from a combination of
two factors. The first factor is a coexistence (or proximity) of different
dynamical regimes near the working point of the model. The second
factor is the presence of temporally correlated noise acting upon the
model, which enforces flips between different dynamical regimes [3].
Here, we examine systematically the effect of this stochastic driving on
the single compartment and on collective effects occurring in the
network of interconnected compartments.
We show that the correlation time of the noise is one of the crucial factors
that modulate the preference of dynamical regimes for a single
compartment, and thus determine the models activity. Furthermore, we
show that this dependence has non-trivial, non-monotonous characteristics.
We find that experimental results obtained in the absence of a stimulus are
best reproduced when the model operates close to a bifurcation. In the
presence of the stimulus, computational results recapitulate the
experimental observations when the alpha resonance is not fully developed.
The emergence of resonance depends on the driving signal amplitude,
properties of the noise, and it relies on collective synchronization, which in
turn depends on the coupling between the network elements. The model
allows us to make predictions for driving with frequencies greater than the
one used in the experiment. Our results indicate that the cross-frequency
transfer observed experimentally may occur only for slow driving: for faster
input the alpha peak does not increase and might even be suppressed.
In summary, our results show that a 1/fb realistic power spectrum with an
embedded alpha peak can be obtained from a neural mass model driven
by a temporally correlated noise. The models dynamics depend in a non-
monotonous way on the correlation time of the noise. Slow oscillatory
driving of this model operating close to a bifurcation facilitates
development of a resonance in the alpha band, giving rise to a cross-
frequency power transfer, which reproduces experimental observations [1].
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Acknowledgements: This work was supported by the European
Commission through the FP7 Marie Curie Initial Training Network 289146
(NETT: Neural Engineering Transformative Technologies), and by the
Ministerio de Economia y Competividad (Spain, project FIS2012-37655).
JGO acknowledges support from the ICREA Academia program.
References
1. Bayer L, Constantinescu I, Perrig S, Vienne J, Vidal P-P, Mhlethaler M,
Schwartz Sophie: Rocking synchronizes brain waves during a short nap.
Curr. Biol 2011, 21:R461-R462.
2. Jansen BH, Rit VG: Electroencephalogram and visual evoked potential
generation in a mathematical model of coupled cortical columns. Biol.
Cybern 1995, 73:357-366.
3. Jedynak M, Pons AJ, Garcia-Ojalvo J: Cross-frequency transfer in a stochastically
driven mesoscopic neuronal model. Front. Comput. Neurosci 2015, 9:1-12.
P76
Modeling thalamic dynamics with a network of integrate and fire neurons
Alessandro Barardi1,2*, Alberto Mazzoni3, Jordi Garcia-Ojalvo2
1
Department of Physics and Nuclear Engineering, Universitat Politecnica de
Catalunya, Terrassa, 08222 Spain; 2Department of Experimental and Health
Sciences, Universitat Pompeu Fabra, Barcelona, Barcelona, 08003 Spain;
3
The BioRobotics Institute, Scuola Superiore SantAnna, Pontedera, 56026, Italy
E-mail: alessandro.barardi@upf.edu
BMC Neuroscience 2015, 16(Suppl 1):P76
Spindle oscillations are rhythmic oscillations at 8-14 Hz originated in the
thalamus during slow-wave sleep. The generation of this rhythm is due to
the rebound bursting properties of thalamo-cortical (TC) relay cells, which
are mutually connected with the thalamic reticular neurons (RE). Several
experiments have shown that during sleep TC neurons can produce bursts
of action potentials due to disinhibition from RE neurons. These bursts in
turn trigger an inhibitory rebound in TC neurons thus giving rise to an
endogenous oscillatory activity [1,2]. Various sophisticated HH-based models
have described in detail the electrophysiological properties of these neurons
and replicate most of the experimental observations made in-vivo and in-
vitro [2]. However these types of models are computationally very
expensive, thus raising the question of whether this dynamical behavior can
be reproduced in simpler models.
To address this question, we used an adaptive exponential integrate and
fire model (aEIF) [3] and studied the conditions under which spindle
oscillations appear in large neuronal networks. The aEIF model can
reproduce the rebound bursting properties of thalamic cells and has been
reported to replicate the self-sustained oscillatory behavior in a minimal 4-
neuron circuit [4]. In this minimal motif, RE cells exhibit periodic sequences
of bursts at 10 Hz (spindles) and recruit TC neurons, which oscillate in anti-
phase at half the frequency (cycle skipping) tightly correlated with IPSP
bursts. However, it is unclear whether these spindle oscillations are
preserved when the neurons are interconnected in a larger network, while
keeping their intrinsic properties.
In order to examine this issue, we have systematically extended the size of
the network of TC and RE neurons, and checked whether the features of
the above-described oscillatory activity hold out. We found the reciprocal
interaction between TC and RE to be essential for sustaining this activity,
coherently with experimental results [5]. Furthermore, some of the RE
neurons exhibit periodic bursting at the spindle frequency, provided they
are not connected to TC neurons directly, but only indirectly through local
inhibitory connections with RE neighbors. Interestingly, this behavior
seems to require that the RE layer has a certain degree of clustering. Under
these conditions, a few TC neurons in the network respond with a burst
every two cycles (cycle skipping). The fact that most TC do not exhibit
cycle skipping could associated with a large strength of GABA IPSPs
impinging on them, according to results of HH kinetic models [6].
In summary, our aEIF network model provides a computationally efficient
description of the dynamical features of the thalamus while preserving
the properties of the individual neurons.
References
1. Destexhe S, Sejnowski: Interactions Between Membrane Conductances
Underlying Thalamocortical Slow-Wave Oscillations. Physiol Rev 2003,
83:1401-1453.
Page 59 of 200
2. Anderson P, Eccles JC: Inhibitory phasing of neuronal discharge. Nature
1962, 196:645-647.
3. Brette R, Gerstner W: Adaptative exponential integrate-and-fire model as
an effective description of neuronal activity. J Neurophysiol 2005,
94:3637-3642.
4. Destexhe A: Self-sustained asynchronous irregular states and Up-Down
states in thalamic, cortical and thalamocortical networks of nonlinear
integrate-and-fire neurons. J Comput Neurosci 2009, 48:1483-1490.
5. Steriade M, Deschenes M, Domich L, Mulle C: Abolition of spindle
oscillations in thalamic neurons disconnected from nucleus reticularis
thalami. J Neurophysiol 1985, 54:1473-1497.
6. Destexhe A, Bal T, McCormick D, Sejnowski T: Ionic mechanisms
underlying synchronized oscillations and propagating waves in a model
of ferret thalamic slices. J Neurophysiol 1996, 76:2049-2070.
P77
Surprise minimization as a learning strategy in neural networks
Mohammad Javad Faraji1*, Kerstin Preuschoff2, Wulfram Gerstner1
1
School of Life Sciences, Brain Mind Institute and School of Computer and
Communication Sciences, Ecole Polytechnique Federal de Lausanne (EPFL),
CH-1015, Lausanne, Switzerland; 2Geneva Finance Research Institute,
University of Geneva, CH-1211, Geneva, Switzerland
E-mail: mohammadjavad.faraji@epfl.ch
BMC Neuroscience 2015, 16(Suppl 1):P77
Surprise is informative because it drives attention and modifies learning.
Not only has it been described at different stages of neural processing [1],
but it is a central concept in higher levels of abstraction such as learning
and memory formation [2]. Several methods, including Bayesian and
information theoretical approaches, have been used to quantify surprise. In
Bayesian surprise, only data observations which substantially affect the
observers beliefs yield surprise [3,4]. In Shannon surprise, however,
observations that are rare or less likely to happen are considered surprising
[5]. Although each of the existing measures partly incorporates conceptual
aspects of surprise, they still suffer from some drawbacks including
implausibility from the view point of neural implementation.
We first review the two probability-based surprise measures above, and
discuss their pros. We then propose a novel measure for calculating surprise
which benefits from the advantages of both measures. Importantly, the
proposed measure benefits from calculating surprise during learning phase
(e.g., inference about parameters in Bayesian framework). This is in contrast
to Bayesian surprise where the surprise calculation is not prior to the
inference step. Our proposed method can also be neurally implemented in a
feed-forward neural network.
Furthermore, we propose a principle of (future) surprise minimization as a
learning strategy; that is if something unexpected (surprising) happens, the
subjective internal model of the external world should be modified such
that the same observation becomes less surprising if it happens again in
the not so distant future. We mathematically describe a class of learning
rules which obey that principle. We show that standard Bayesian updating
and the likelihood maximization technique both belong to such class. It
accredits usage of well-known inference techniques in frequentist and
Bayesian frameworks from a novel perspective. As a consequence, we
propose a modified Bayesian method for updating beliefs about the world.
This learning rule also obeys the principle of surprise minimization. In this
method, the influence of the likelihood term on the posterior belief can be
controlled by a subjective parameter. We apply this technique to learning
within changing environments. Modified Bayesian updating helps the
learning agent to actively control the influence of new information on
learning environments. As a result, the agent quickly adapts to the
changing environments.
Acknowledgements: This research was supported by the European
Research Council (grant agreement no. 268 689).
References
1. Fairhall AL, Lewen GD, Bialek W, van Steveninck RRR: Efficiency and
ambiguity in an adaptive neural code. Nature 2001, 412(6849):787-792.
2. Ranganath C, Rainer G: Neural mechanisms for detecting and
remembering novel events. Nature Reviews Neuroscience 2003,
4(3):193-202.
3. Baldi P, Itti L: Of bits and wows: a Bayesian theory of surprise with
applications to attention. Neural Networks 2010, 23(5):649-666.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
4. Itti L, Baldi P: Bayesian surprise attracts human attention. Advances in
neural information processing systems 2005, 547-554.
5. Shannon CE: A mathematical theory of communication. ACM SIGMOBILE
Mobile Computing and Communications Review 2001, 5(1):3-55.
P78
A hierarchy of time scales supports unsupervised learning of behavioral
sequences
Samuel P Muscinelli*, Wulfram Gerstner
School of Life Sciences, Brain Mind Institute and School of Computer and
Communication Sciences, cole Polytechnique Fdrale de Lausanne, 1015
Lausanne, Switzerland
E-mail: samuel.muscinelli@epfl.ch
BMC Neuroscience 2015, 16(Suppl 1):P78
Playing the piano, speaking or playing tennis are just few examples of
behavioral situations in which we need to perform sequences of actions.
The neurophysiological mechanisms that underlie the production and
learning of such sequences are far from being understood and several
crucial issues arise in biologically plausible models: first, there is a huge
gap of time scales between the response of single neurons, which is on
the order of tens of milliseconds, and common behavioral situations,
which usually span several seconds. Moreover, behavioral sequences are
often complex, i.e. they cannot be described as Markov chains and thus
require to relate distant parts of the sequence. Finally, similar to its
biological counterpart, a candidate model should be able to learn new
sequences with biologically plausible learning rules.
We devise a simplified model of neural populations with the aim of
producing slow sequences of neural activation to correlate with behavioral
sequences. We exploit spike frequency adaptation of single neurons to
introduce a slow process [1] in the population dynamics in order to fill the
gap between different time scales described above. Our model features a
hierarchy of time scales inspired by evidence of different time scales among
different areas in the brain [2,3]. This allows a separation of neural coding
into varying levels of temporal detail: the sequence produced in a faster
area will be regulated by the activity of slower areas in a hierarchical
fashion. This provides a mechanism to deal with non-Markovianity thanks to
the longer memory capacity of slower areas. Finally, we show that it is
possible to learn the appropriate inter-area synaptic connections using
biologically plausible learning rules, exploiting the decrease in activity due
to adaptation. This approach leads to the development of temporal
receptive fields associated to subparts of the desired sequence with
hierarchical levels of detail.
Our model constitutes a promising approach to temporal learning,
showing how an appropriate neural substrate with a hierarchy of time
scales can lead, even without any error or reward signal, to the learning
of slow and complex sequences.
Acknowledgements: Research was supported by the Swiss National
Science Foundation (grant agreement no. 200020_147200).
Figure 1(abstract P79) Stochastic population equation precisely captures the collective bistable dynamics of a spiking neural network. A Two
mutually inhibitory populations of 500 neurons each. B,C Sample paths of the spiking neural network and the derived population model, respectively.
Transitions are due to finite-size fluctuations. D The population activity of microscopic and mesoscopic model have the same first- and second-order statistics
Page 60 of 200
References
1. Pozzorini C, Naud R, Mensi S, Gerstner W: Temporal whitening by power-
law adaptation in neocortical neurons. Nat Neurosci 2013, 16(7):942-948.
2. Kiebel SJ, Daunizeau J, Friston KJ: A hierarchy of time-scales and the
brain. PloS Comput Biol 2008, 4(11):e1000209.
3. Hasson U, Yang E, Vallines I, Heeger D, Rubin N: A hierarchy of temporal
receptive windows in human cortex. J Neurosci 2008, 28(10):2539-2550.
P79
Bridging spiking neuron models and mesoscopic population models - a
general theory for neural population dynamics
Tilo Schwalger*, Moritz Deger, Wulfram Gerstner
Brain Mind Institute, cole Polytechnique Fdrale de Lausanne (EPFL)
Station 15, 1015 Lausanne, Switzerland
E-mail: tilo.schwalger@epfl.ch
BMC Neuroscience 2015, 16(Suppl 1):P79
Many circuits of the brain can be described by a system of interacting
neural populations that are approximately homogeneous. For instance,
cortical layers typically consist of a few main types of excitatory and
inhibitory neurons that form small homogeneous populations of neurons.
Such systems can be modeled on different spatial scales. On the
microscopic scale, single cell activity has been faithfully described by
reduced phenomenological neuron models [1]. Simulations of networks of
such neuron models are, however, computationally expensive and do not
offer much analytical insight. On the other hand, mesoscopic population
models are reduced descriptions of the global activities of each population.
These activities are stochastic due to the finite sizes of the populations.
Mesoscopic models can be efficiently simulated and provide a better
understanding of the dynamics owing to the abstraction of microscopic
information. However, it is largely unknown how to relate mesoscopic
population models to microscopic properties such as neural refractoriness,
synaptic conductance dynamics and spike-frequency adaptation.
Here, we derive a mesoscopic population model for microscopic networks
of generalized integrate-and-fire neurons [1]. This type of neuron model
supports important properties like neural refractoriness, multiple-time-
scale adaptation, stochastic spike generation and synaptic dynamics; its
parameters can be directly extracted from experiments of real neurons. In
particular, we use a mean-field and a quasi-renewal approximation [1] to
derive stochastic integral equations for the population rates. These
equations highlight how the history of activities and fluctuations affects
the refractoriness of the populations and the activities at the current
time. They can be solved forward in time and thus allow to quickly
generate stochastic samples of spontaneous or evoked activities (Fig. 1B,
C) that have the same statistics as a corresponding microscopic network
simulation to a high degree of accuracy (Fig. 1D). The theory not only
captures linear population dynamics [2] but also nonlinear collective
effects that emerge on the population level such as metastability (Fig. 1).
Our novel theory establishes a general framework for modeling neural
population dynamics based on microscopic neuronal parameters. It offers
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
an efficient way to analyze cortical circuits and its computational
functions, and how they depend on single-cell and synaptic properties.
Acknowledgements: Research was supported by the European Research
Council (Grant Agreement no. 268689, MultiRules).
References
1. Gerstner W, Kistler WM, Naud R, Paninski L: Neuronal Dynamics: From Single
Neurons to Networks and Models of Cognition Cambridge: Cambridge
University Press 2014.
2. Deger M, Schwalger T, Naud R, Gerstner W: Fluctuations and information
filtering in coupled populations of spiking neurons with adaptation. Phys
Rev E 2014, 90(6-1):062704.
P80
Variable bin size selection for periestimulus time histograms (PSTH)
with minimum mean square error criteria
SM Heidarieh1*, M Jahed1, A Ghazizadeh2
1
Electrical Engineering, Sharif University of Technology, Tehran, Iran; 2Laboratory
of Sensorimotor Research, National Institutes of Health, Bethesda, MD, USA
E-mail: heidarieh@ee.sharif.edu
BMC Neuroscience 2015, 16(Suppl 1):P80
To date the most common method for extracting neuronal responses is
by constructing PSTHs that are time locked to task events. Many
parameters of interests such as response magnitude, onset and duration
are then calculated from the constructed PSTHs. However the precision of
PSTH response estimate critically depends on the choice of bin sizes. This
dependence demands objective criteria for bin size selection. A seminal
study by Shimazaki and Shinomoto [1] derived an optimal cost function
for choosing a fixed bin size for a time varying Poisson process. It is easy
to see that using a one-size-fit-all recipe for bin sizes will invariably
overestimate and underestimate rate changes for fast and slow
fluctuations respectively.
Here we extend previous results by calculating the cost function that
minimizes mean square error for variable bin sizes with the same
assumptions used previously for time varying Poisson processes
 
1 N 2
2ki (ki i k)
Cost(N, ) = 2 . To
n T i=1 i operations. After suppression of strong external luminance gradients
minimize this nonlinear and nonconvex cost function, we utilize an
Figure 1(abstract P80) Comparing the costs and time efforts of the
algorithms
Page 61 of 200
array of methods some of which are widely used for nonlinear
optimization, namely: Active set, Simultaneous perturbation stochastic
approximation (SPSA), Genetic Algorithm and an approximate heuristic
algorithm. Average performance of each algorithm on a typical
simulated neuronal firing is calculated using 50 iterations. All methods
resulted in a lower cost function compared to fixed bin size as
expected. Plotting the final cost vs the algorithm run time shows that the
method of Active set overall has the best cost reduction while still being
reasonably fast compared to the fixed bin size approach (Figure 1) .
Further investigation of the properties of this cost function and
developing computationally efficient methods for its minimization will be
the basis of future work.
Reference
1. Shimazaki H, Shinomoto S: A method for selecting the bin size of a time
histogram. Neural computation 2007, 19(6):1503-1527.
P81
Neural model of biological motion recognition based on shading cues
Leonid A Fedorov*, Martin A Giese*
Section f. Computational Sensomotorics, Dept. of Cogn. Neurology, CIN/ HIH,
University Clinic Tuebingen, Tuebingen, Germany
E-mail: leonid.fedorov@cin.uni-tuebingen.de; martin.giese@uni-tuebingen.de
BMC Neuroscience 2015, 16(Suppl 1):P81
Point-light or stick-figure biological motion stimuli, due to the absence of
depth cues, can induce bistable perception, where the walker is
perceived as heading in two alternating directions [1,2]. Psychophysical
studies suggested an importance of depth cues for biological motion
perception [3]. However, neural models of biological motion perception
so far have focused on the processing of features that characterize the
2D structure and motion of the human body [4,5]. We extend such
models for the processing of shading cues in order to analyze the three-
dimensional structure of walkers from monocular stimuli.
Model: As extension of a learning-based neural model [4], we add a
shading pathway that computes the internal contrast gradients that vary
with the 3D view of the walker, even if the silhouette information remains
identical (Figure 1A-C). The model exploits physiologically plausible
caused by the boundaries of the silhouette, internal luminance gradient
features are extracted by a hierarchy of neural detectors. These gradient
features, combined with the shape features extracted by the form pathway
of the model in [4], are used as input for snapshot neurons, RBF units that
detect 3D body shapes (Figure 1D). These model neurons are embedded
within a two-dimensional recurrent neural field [6] that jointly represents
the sequential temporal structure of the stimulus and the view of the
walker.
Results: The neural field dynamics reproduces perceptual multi-stability
and spontaneous perceptual switching between stimulus views, observed
for silhouette stimuli in psychophysical experiments [1,2]. It also
reproduces the disambiguation by addition of shading information and a
new perceptual illusion, which illustrates a lighting-from-above prior in
the processing of biological motion stimuli.
Acknowledgements: Supported by EC FP7 ABC PITN-GA-011-290011, HBP
FP7-604102, Koroibot FP7-611909, COGIMON H2020-644727, DFG GI 305/4-1,
DFG GZ: KA 1258/15-1, and BMBF, FKZ: 01GQ1002A.
References
1. Vanrie J, Dekeyser M, Verfaillie K: Bistability and biasing effects in the
perception of ambiguous point-light walkers. Perception 2004,
33:547-560.
2. Vangeneugden J, De Mazire P, Van Hulle M, Jaeggli T, Van Gool L,
Vogels R: Distinct mechanisms for coding of visual actions in macaque
temporal cortex. J Neurosci 2011, 31(2):385-401.
3. Vanrie J, Verfaillie K: Perceiving depth in point-light actions. Perc
Psychophys 2006, 68(4):601-612.
4. Giese MA, Poggio T: Neural mechanisms for the recognition of biological
movements and action. Nat Rev Neurosci 2003, 4:179-192.
5. Lange J, Lappe M: A model of biological motion perception from
configural form cues. J Neurosci 2006, 26:2894-2906.
6. Amari S: Dynamics of pattern formation in lateral inhibition type neural
fields. Biol Cyb 1977, 27:77-87.
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 62 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

Figure 1(abstract P81) A. Snapshot from a walker stimulus, rendered from a -45 side view. Vectors indicate internal luminance gradients, extracted
by the internal gradient detectors of the model. B. Silhouette stimulus without shading cues is ambiguous and compatible with view angles 45. C.
Snapshot and internal shading gradients for +45 side view. D. Shading pathway. After suppression of strong boundary gradients, internal luminance
gradients are extracted, using a hierarchy of neural detectors similar to a convolutional network. At the highest level is formed by snapshot neurons, RBF
units that have been trained with keyframes from 3D walker movies, which are embedded in a dynamic neural field

accessible formats is an important strategy for improving transparency and

P82
Visualizing, editing and simulating neuronal models with the Open
Source Brain 3D explorer
Adrian Quintana1, Matteo Cantarelli2, Boris Marin1, R Angus Silver1,
Padraig Gleeson1*
1
Department of Neuroscience, Physiology and Pharmacology, University
College London, London, UK; 2Metacell LLC, San Diego, CA, USA
E-mail: p.gleeson@ucl.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P82
Reproducibility, accessibility, standardization, and provenance tracking are
some of the main challenges for the global scientific community today. In
computational neuroscience, making models available in open and
reusability of published models. The Open Source Brain initiative (OSB,
www.opensourcebrain.org) is an online platform which aims to facilitate
sharing and collaborative development of neuronal models. It provides a
central location for models of multiple brain regions and species and a set
of tools to analyze, edit, visualize and simulate them. While OSB can host
models in any format, converting models to NeuroML (specifically v2.0, [1])
is actively encouraged as it allows OSB to access the internal dynamics of
the model. NeuroML2 is a simulator independent model description
language for computational neuroscience. It has been built on LEMS (Low
Entropy Model Specification, [1]), a general purpose language developed
to provide a machine readable way of expressing the structure and
dynamics of physical models. These languages facilitate cross-simulator

Figure 1(abstract P82)

BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
validation, which is a crucial factor in ensuring reproducibility and
reusability of models.
OSB currently hosts more than 70 projects, has 350 users, and 40 member
labs. The vast majority of its features are accessible to non-registered users.
If a specific model has been converted to NeuroML2/LEMS the user can
make use of a wide range of advanced features in the Open Source Brain
3D explorer. The OSB 3D explorer is a graphical framework for analysis and
simulation of neural models based on Geppetto (http://www.geppetto.
org). Geppetto, an open source modular platform for complex biological
systems, was originally developed as part of the Open Worm project
(http://www.openworm.org, [2]). OSB 3D explorer provides a 3D canvas in
order to display the model neurons/network. In addition, a wide variety of
widgets allow the user to access data on both the model and the
simulation. A Tree Visualiser widget allows the user to analyze the
dynamics of the model, visualize a summary and to modify aspects of the
model. The model can be downloaded in different formats (such as
NEURON, Brian, Matlab, etc.) or can be simulated on a server/cluster. Such
simulations will be executed asynchronously and the user can manage the
simulations and analyze the results through a web dashboard. Some other
widgets provide tools to analyze the connections in the network, inspect
variables during the simulation or generate plots showing for instance the
dynamics of a gate or the membrane potential of a cell during the
simulation. Every change in the model and simulation outputs is stored in
the database providing the user traceability and reproducibility of the
whole experiment.
Acknowledgements: This work has been primarily funded by the
Wellcome Trust (101445/095667). BM is supported by the Brazilian agency
CAPES (2257-13/0).
References
1. Cannon RC, Gleeson P, Crook S, Ganapathy G, Marin B, Piasini E, Silver RA:
LEMS: a language for expressing complex biological models in concise
and hierarchical form and its use in underpinning NeuroML2. Frontiers in
Neuroinformatics 2014, 8:79.
2. Szigeti B, Gleeson P, Vella M, Khayrulin S, Palyanov A, Hokanson J, Currie M,
Cantarelli M, Idili G, Larson S: OpenWorm: an open-science approach to
modelling Caenorhabditis elegans. Front. Comput. Neurosci 2014.
P83
Effects of spike-time dependent plasticity on deep brain stimulation of
the basal ganglia for treatment of Parkinsons disease
Logan L Grado1*, Matthew D Johnson1,2, Theoden I Netoff1
1
Graduate Program in Biomedical Engineering, University of Minnesota,
Minneapolis, MN 55455, USA; 2Institute for Translational Neuroscience,
University of Minnesota, Minneapolis, MN 55455, USA
E-mail: grado010@umn.edu
BMC Neuroscience 2015, 16(Suppl 1):P83
Deep brain stimulation (DBS) of the basal ganglia is a widely used and
effective treatment for patients with medication-refractory Parkinsons
disease (PD). However, tuning the stimulation parameters to maximize
therapy while minimizing side effects is performed mostly through a trial
and error approach, consuming time and energy of both clinician and
patient [1]. As such, there is a need for a systematic, engineering based
approach to improve patient outcomes. Current theories of DBS
mechanisms propose that DBS suppresses pathological oscillations (15-35
Hz) that dominate the basal ganglia by suppressing information flow [2,3].
However, effects of DBS are not instant, often taking minutes or more
before an effect is seen [4]. This time scale suggests DBS may induce a
change in network architecture through synaptic plasticity, destabilizing
oscillations in the network. Recently, a new approach to DBS called
Coordinated Reset appears to take advantage of this phenomenon,
resulting in therapeutic benefits that last from hours to days [5,6]. We
hypothesize that if beta oscillations are indeed responsible for
Parkinsonian signs, then the dissipation and return of these oscillations
should follow a time course similar to that of the symptoms themselves.
We use a computational network model of PD with emergent pathological
34 Hz oscillation developed by Hahn & McIntyre [7] to test the effects of
DBS on the basal ganglia, implemented with spike-time dependent
plasticity (STDP) as described by Badoual et al. [8]. Preliminary results show
that with the introduction of STDP, pathological beta oscillations dissipate
over time after the onset of DBS stimulation in computational models. This
Page 63 of 200
work suggests that it may be possible to tune stimulation settings to take
advantage of long term plasticity effects in DBS to improve patient
outcomes.
Acknowledgements: Research supported by Neuroengineering NSF
IGERT under DGE-1069104.
References
1. Volkmann J, Moro E, Pahwa R: Basic algorithms for the programming of
deep brain stimulation in Parkinsons disease. Mov Disord 2006,
21(Suppl 1):S284-S289.
2. Johnson MD, Miocinovic S, Mcintyre CC, Vitek JL: Mechanisms and Targets
of Deep Brain Stimulation in Movement Disorders 2008, 5(April):294-308.
3. Agnesi F, Connolly AT, Baker KB, Vitek JL, Johnson MD: Deep brain
stimulation imposes complex informational lesions. PLoS One 2013, 8:
e74462.
4. Cooper SE, Noecker AM, Abboud H, Vitek JL, McIntyre CC: Return of
bradykinesia after subthalamic stimulation ceases: relationship to
electrode location. Exp Neurol 2011, 231:207-13.
5. Tass P a, Qin L, Hauptmann C, Dovero S, Bezard E, Boraud T, Meissner WG:
Coordinated reset has sustained aftereffects in Parkinsonian monkeys.
Ann Neurol 2012, 72:816-20.
6. Adamchic I, Hauptmann C, Barnikol UB, Pawelczyk N, Popovych O,
Barnikol TT, Silchenko A, Volkmann J, Deuschl G, Meissner WG, Maarouf M,
Sturm V, Freund H-J, Tass PA: Coordinated reset neuromodulation for
Parkinsons disease: Proof-of-concept study. Mov Disord 2014, 29:1679-84.
7. Hahn PJ, McIntyre CC: Modeling shifts in the rate and pattern of
subthalamopallidal network activity during deep brain stimulation.
J Comput Neurosci 2010, 28:425-41.
8. Badoual M, Zou Q, Davison AP, Rudolph M, Bal T, Frgnac Y, Destexhe A:
Biophysical and phenomenological models of multiple spike interactions
in spike-timing dependent plasticity. Int J Neural Syst 2006, 16:79-97.
P84
Computational modelling predicts activity-dependent neuronal
regulation by nitric oxide increases metabolic pathway activity
Christophe B Michel1*, Sarah J Lucas2, Ian D Forsythe2, Bruce P Graham1
1
Computing Science and Mathematics, University of Stirling, Stirling,
Scotland, FK9 4LA, UK; 2Dept Cell Physiology & Pharmacology, University of
Leicester, Leicester LE1 9HN, UK
E-mail: cmi@cs.stir.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P84
The neuromodulator nitric oxide (NO), in addition to regulating
electrophysiological homeostasis, post-synaptic receptor activity, and
neuronal functions through cyclic GMP, has recently been shown to
modulate the metabolic pathway. NO down-regulates mitochondrial
activity [1] and the subsequent increase in AMP facilitates the activation of
the phosphofructokinase enzymatic reaction in the glycolytic pathway [2].
Given these results, we want to better understand the regulation of neuronal
energy metabolism by NO. To do that, we have built a computational model
of energy metabolism based on prior works, principally on a model
elucidating the control system structures of neuronal metabolism [3] and
another highlighting the metabolic role between neuronal activity and
hemodynamics [4]. From the first we took the biochemical pathway model,
i.e. glycolysis, mitochondrial activity and regulation by the astrocyte to
neuron lactate shuttle. From the second we took a model of intracellular
sodium concentration and pumping by Na/K-ATPase. We added the activity
dependent glutamate cycle [5], driven by electrophysiological activity.
We finally completed these models with the previously described
modulation by NO.
We have new experimental recordings of post synaptic currents in principal
cells of the mouse medial nucleus of the trapezoid body (MNTB) in response
to high frequency presynaptic stimulation. To match our model to this data
we assumed the neurotransmitter released during electrophysiological
activity to be proportional to the post synaptic current amplitude and so our
glutamate cycle synaptic model, described above, could be related to the
recorded EPSC amplitudes. The full metabolism model was then calibrated
to match actual voltage clamp MNTB recordings of synapse activity, in the
control case and in the presence of 5 mM 2-deoxy-D-glucose, blocking
glycolysis, allowing only mitochondrial activity.
Model outputs are compared between (1) the baseline (control)
conditions, (2) following conditioning with evoked activity that increases
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
NO levels, modifying individually glycolysis and mitochondrial activity,
and (3) in full NO regulated conditions, in order to evaluate the individual
and mixed metabolism dynamics. Results show that ATP production is
slightly increased by NO upregulation of glycolysis (2.5%), significantly
increased by mitochondrial inhibition by NO (12.3%) and further
increased (15.6%) when both glycolysis and mitochondria are NO
modulated.
This model will eventually be combined with our previous postsynaptic
model that shows NO modulation can reduce the cost of action potential
generation in MNTB neurons [6] to build a more complete model of
energy consumption during synaptic transmission.
Acknowledgements: This work was funded by BBSRC grant BB/K01854X/
1 to BPG and BB/K01899X/1 to IDF.
References
1. Erusalimsky JD, Moncada S: Nitric oxide and mitochondrial signaling: from
physiology to pathophysiology. Arterioscler Thromb Vasc Biol 2007,
27(12):2524-2531.
2. Bolaos JP, Delgado-Esteban M, Herrero-Mendez A, Fernandez-Fernandez S,
Almeida A: Regulation of glycolysis and pentose-phosphate pathway by
nitric oxide: impact on neuronal survival. Biochim Biophys Acta 2008,
1777(7-8):789-793.
3. Aubert A, Costalat R: A model of the coupling between brain electrical
activity, metabolism, and hemodynamics: application to the
interpretation of functional neuroimaging. NeuroImage 2002,
17(3):1162-1181.
4. Cloutier M, Wellstead P: The control systems structures of energy
metabolism. J. R. Soc. Interface R. Soc 2010, 7(45):651-665.
5. Cloutier M, Bolger FB, Lowry JP, Wellstead P: An integrative dynamic
model of brain energy metabolism using in vivo neurochemical
measurements. J. Comput. Neurosci 2009, 27(3):391-414.
6. Michel CB, Hennig MH, Graham BP: Nitric oxide activity-dependent regulator
compensates synaptic depression and enhances metabolic efficiency in
the auditory brainstem. BMC Neuroscience 2014, 15(Suppl 1):P154.
P85
Computational neural modelling of auditory cortical receptive fields
Jordan D Chambers1*, Anthony N Burkitt1,2, David B Grayden1,2,3
1
NeuroEngineering Laboratory, Department of Electrical and Electronic
Engineering, University of Melbourne, Parkville, Victoria 3010, Australia;
2
Bionics Institute, East Melbourne, Victoria 3002, Australia; 3Centre for Neural
Engineering, University of Melbourne, Parkville, Victoria 3010, Australia
E-mail: jordanc@unimelb.edu.au
BMC Neuroscience 2015, 16(Suppl 1):P85
Introduction: Acoustic signals are mainly characterized by their temporal
dynamics. Electrophysiological studies have shown that neurons in the
primary auditory cortex (A1) can detect fine temporal structure of acoustic
stimuli [1]. The traditional view of auditory processing describes how the
temporal codes of sounds are distributed in the frequency domain along
the auditory pathway from the basilar membrane to the cortex. Therefore,
it is important to consider a sounds spectral and temporal features
together. The spectrotemporal receptive field (STRF) is a description of the
auditory systems input-to-output transformation encompassing both the
spectral and temporal features. The STRF of A1 neurons exhibit complex
patterns that can undergo rapid task-related changes [2]. However, the
mechanisms by which cortical neurons change their STRF remains unclear.
Methods: A computational neural model was developed to investigate
mechanisms by which cortical neurons can change their STRF. A sound
signal is played into a model of the cochlear and cochlear nucleus (CN) [3].
It comprises of a bank constant-Q bandpass filters spread along the
frequency axis, a high-pass filter, a non-linear compression, a low-pass
filter, a first-order derivative with respect to the tonotopic axis, a half-wave
rectifier and integration over a 8 ms window. The output of the CN tuned
to a particular frequency drives the input to integrate-and-fire neuron
models to represent cortical neurons. Action potentials (AP) in the cortical
neurons are then used to calculate a reverse autocorrelation of the STRF.
Results: When white Gaussian noise is used as the sound signal, a region
of STRF excitation is produced with a frequency corresponding to the CN
output. This STRF excitatory region can be increased in the frequency
domain by increasing the number of CN outputs exciting the cortical
neuron. If there is a large separation along the frequency axis, multiple
Page 64 of 200
STRF excitatory regions are created. The excitatory region can be altered
in the temporal domain by adding delays to the CN output. Regions of
STRF inhibition can also be produced when a cortical neuron fires APs
tonically and the CN output hyperpolarizes the cortical neuron.
Alternatively, a region of STRF inhibition can be created without tonic
firing, if one CN output excites the cortical neuron and a different CN
output hyperpolarizes the cortical neuron. This creates a region of
excitation and a region of inhibition in the STRF, similar to complex STRFs
observed experimentally. STRF inhibitory regions can be altered in the
frequency and temporal domains in the same way as excitatory regions.
Changing the properties of the cortical neuron to include an after-
hyperpolarizing potential following an AP causes a region of excitation to
be preceded by a region of inhibition, and a region of inhibition to be
preceded by a region of excitation, hereby providing a secondary means
to producing complex STRFs.
Conclusions: This computational neural model of STRF in A1 can
reproduce experimentally observed complex STRFs. It can predict changes
in synaptic connectivity and AP firing properties occurring in A1 neurons
corresponding to rapid task-dependent changes in the STRFs.
References
1. Elhilali M, Fritz JB, Klein DJ, Simon JZ, Shamma SA: Dynamics of precise
spike timing in primary auditory cortex. Journal of Neuroscience 2004,
24(5):1159-1172.
2. Fritz J, Elhilali M, Shamma S: Active listening: Task-dependent plasticity of
spectrotemporal receptive fields in primary auditory cortex. Hearing
Research 2005, 206(1-2):159-176.
3. Ru PW, Chi TS, Shamma S: The synergy between speech production and
perception. Journal of the Acoustical Society of America 2003, 113(1):498-515.
P86
The interaction between integration and segmentation neurons for
motion perception
Parvin Zarei1,2*, Tatiana Kameneva1, Michael R Ibbotson3, Anthony N Burkitt1,4,
David B Grayden1,4
1
NeuroEngineering Lab, Dept Electrical & Electronic Engineering, University
of Melbourne, Melbourne, VIC 3010, Australia; 2NICTA, Sydney, Australia;
3
National Vision Research Institute, Australian College of Optometry,
Melbourne, VIC 3053, Australia; 4Bionics Institute, Melbourne, VIC 3002,
Australia
E-mail: parvinz@student.unimelb.edu.au
BMC Neuroscience 2015, 16(Suppl 1):P86
The main challenge for understanding motion processing in the visual
system is that local motion signals do not accurately represent the
direction of the whole object. This is known as the aperture problem. The
small receptive fields of individual neurons can only measure motion
components orthogonal to their preferred direction. Neuro-physiological
studies show that neurons in brain area MT (V5) have a major role in
dealing with this problem [1]. We develop a model to overcome the
aperture problem through excitatory and inhibitory interconnections
between MT neurons.
The model contains two levels of rate-based neuron. The first level is a
model of V1 with two sets of neurons: complex neurons and hyper-
complex end-stopped neurons. End-stopped neurons respond to the end-
points of the stimulus , thereby representing an unambiguous estimation
of the direction of motion [2]. The response of V1 neurons is transmitted
to the second level, the model of MT neurons. MT neurons are categorized
as integration and segmentation neurons. The role of integration neurons
is to integrate local motion signals to solve the aperture problem; the role
of segmentation neurons is to detect discontinuities in the input stimulus
to inhibit integration neurons [3].
Excitatory recurrent connections between integration neurons propagate
end-stopped information from the object terminators. As each spatial
location contains neurons selective to different directions, inhibitory
connections are added between the neurons such that there will be
only one neuron active at each spatial location. Long-range inhibitory
connections are also added between more distant MT neurons selective to
other directions. Segmentation neurons send inhibitory connections to
integration neurons to prevent the propagation of motion signals outside
the border of the input stimulus. Inhibitory connections between
segmentation neurons are defined based on center-surround interactions;
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
this suppresses activity when the motion signals in the center and
surround are in the same directions; ensuring that their activity will not
prevent the propagation of motion signals when there is no discontinuity.
The results of simulations show that the activity of V1 neurons along the
edges is much stronger than the activity of the neurons at the end-
points. Activation of end-stopped neurons in V1 suppresses ambiguous
motion signals along the edges of the stimulus. Therefore, unambiguous
information at the end-points wins the competition over the neurons
with ambiguous motion signals. The final output of the model shows that
MT neurons achieve an accurate estimation of motion using the
information provided by the end-stopped neurons.
We developed a model of MT neurons for motion perception based on the
available neuro-physiological data. These neurons receive their input from
neurons in V1. The results show the successful function of these neurons in
the perception of motion of a single moving bar. In this model, we also
investigated the role of end-stopped neurons in overcoming the aperture
problem. The results suggest that MT neurons are not able to solve the
aperture problem without the vital role played by end-stopped cells.
Acknowledgements: The Bionics Institute acknowledges the support it
receives from the Victorian Government through its Operational Infra-
structure Support Program. This research was supported by the Australian
Research Council through its Special Research Initiative in Bionic Vision
Science and Technology grant to Bionic Vision Australia.
References
1. Born R, Bradley D: Structure and function of visual area MT. Ann. Rev.
Neurosci 2005, 28:157-189.
2. Tsui JMG, Hunter JN, Born R, Pack C: The role of V1 surround suppression
in MT motion integration. J. Neurophy 2010, 103(6):3123-3138.
3. Liden L, Pack C: The role of terminators and occlusion cues in motion
integration and segmentation: a neural network model. Vis. Res 1999,
39(19):3301-3320.
P87
A simple effective model for STDP: from spike pairs and triplets to
rate-encoding plasticity
Rodrigo Echeveste*, Claudius Gros
Institute for Theoretical Physics, Goethe University Frankfurt, Hessen, 60438,
Germany
E-mail: echeveste@itp.uni-frankfurt.de
BMC Neuroscience 2015, 16(Suppl 1):P87
In the present work [1] we propose an effective model formulating
synaptic potentiation and depression in terms of two interacting traces,
representing the fraction of open NMDA receptors and the Ca2+
Figure 1(abstract P87) Models prediction and comparison to experimental results from hippocampal culture. A . The standard pairwise STDP
curve is recovered by the model. Blue lines indicate the models results and red circles the experimental data [2]. B . Triplets, consisting of two pre- and
one postsynaptic spike or vice versa, induce a non-linear change in synaptic strength. Blue bars represent the models results, which follow closely the
experimental results [3] presented with red boxes. In green, the linear addition of the contribution of the two composing pairs as from Panel A
Page 65 of 200
concentration in the post-synaptic neuron, respectively. These two traces
then determine the evolution of the synaptic strength. We first confirm
that the standard pairwise STDP curve is obtained for low frequency trains
of pairs of pre- and post-synaptic spikes and we then evaluate triplet
effects (see Figure 1), comparing the models results to experimental data
from hippocampal culture [2,3]. Finally, we evaluate the models
predictions for spike trains of different frequencies and degrees of
correlation, observing that a BCM-like rule for plasticity as a function of the
pre-and post-synaptic firing rates is recovered when employing
uncorrelated poisson trains of pre- and postsynaptic spikes.
Having a low number of parameters and being composed of only
polynomial differential equations, the model is able nonetheless to
reproduce key features of LTP and LTD. Moreover, since the parameters of
the model are easily related to the dynamical properties of the synapse,
we believe the model constitutes a useful tool to study extended neural
networks from a dynamical systems point of view.
Acknowledgements: The support of the German Science Foundation (DFG)
and the German Academic Exchange Service (DAAD) are acknowledged.
References
1. Echeveste R, Gros C: Two-trace model for spike-timing-dependent
synaptic plasticity. Neural comput in press.
2. Bi GQ, Poo MM: Synaptic Modifications in Cultured Hippocampal
Neurons: Dependence on Spike Timing, Synaptic Strength, and
Postsynaptic Cell Type. J Neurosci 1998, 18:10464-10472.
3. Wang HX, Gerkin RC, Nauen DW, Bi GQ: Coactivation and timing-dependent
integration of synaptic potentiation and depression. Nat Neurosci 2005,
8:87-193.
P88
Slow points and adiabatic fixed points in recurrent neural networks
Hendrik Wernecke*, Claudius Gros
Institute for Theoretical Physics, Goethe-University, Frankfurt am Main, 60438,
Germany
E-mail: wernecke@itp.uni-frankfurt.de
BMC Neuroscience 2015, 16(Suppl 1):P88
The time scales for cognitive information processing in the brain range, at
least, from milliseconds (the time scale of the action potential) to many
seconds (the time scale of short-term memory), spanning several orders of
magnitude. In this context the slow dynamical components can be regarded
as background processes modulating adiabatically the parameters
governing the fast neural activity. For the case of recurrent neural networks
the slow processes then change adiabatically the attractor landscape,
including the adiabatic fixed points, inducing possibly both second order
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P88) Phase diagram of the three-neuron system showing five different phases distinguished by the stability and the shape
of the adiabatic fixed points (AFP) for different values of the adaption rate . One can observe three second order transitions (dashed) and one
first order transition showing hysteresis (striped area). The subplots show the firing rate (green), saddle node AFP (gray), stable AFP (blue) and effectively
attracting AFP (red) over time and represent a phase each
bifurcations, with respect to the steady-state neural activity, and first order
catastrophes.
In this contribution we investigate the slow adaption of the attractor
landscape of exemplary small recurrent neural networks consisting of
continuous-time point neurons [1]. The state of one of these integrate-
and-fire neurons is fully determined by its membrane potential and two
adapting internal parameters [2], the gain and the threshold, with the
time scale of adaption 1/ being substantially larger than the time scale
of the primary neural activity. We point out that not only the adiabatic
fixed points of the network are important for shaping the neural
dynamics, but also the points in phase space where the flow slows down
considerably (called slow points or attractor ruins [3]).
We rigorously examine the metadynamics of the attractor landscape for a
three-neuron system, observing five different phases (see Fig. 1) for
different values of the adaption rate, of which four can be distinguished
by the number and stability of the adiabatic fixed points. Three of the
observed transitions are of second order. The remaining transition to the
phase of lowest adaption is of first order and shows hysteresis. This
transition occurs, remarkably, at very low adaption rates of ~ 10-5 and is
characterized by a higher-order catastrophe.
We conclude that even relatively simple recurrent networks may show
highly non-trivial adapting attractor landscapes and that the study of the
attractor metadynamics in the brain maybe important for a further
understanding of decision processes and dynamical memory recall.
Acknowledgements: This work benefited from discussions with Bulcs
Sndor. The research was supported by funds of the DFG and Studienstiftung
d. dt. Volkes.
References
1. Linkerhand M, Gros C: Generating functionals for autonomous latching
dynamics in attractor relict networks. Sci Rep 2013, 3.
2. Triesch J: A gradient rule for the plasticity of a neurons intrinsic
excitability. Proceedings of ICANN Springer 2005, 65-70.
3. Gros C, Linkerhand M, Walther V: Attractor Metadynamics in Adapting
Neural Networks. ArXiv preprint 2014, arXiv:1404.5417.
P89
Limit cycles with transient state dynamics in cyclic networks
Bulcs Sndor*, Claudius Gros
Institut fr Theoretische Physik, Goethe Universitt, Frankfurt am Main, 60438,
Germany
E-mail: sandor@itp.uni-frankfurt.de
BMC Neuroscience 2015, 16(Suppl 1):P89
Changes in the transmission properties of synapses may influence actively
the processing of information. This is in particular the case for the working
Page 66 of 200
memory, viz the temporary storage of information, which is thought to be
mediated via short-term or transient synaptic plasticity effects [1]. The
standard Tsodyks-Markram model [2] for short-term synaptic plasticity
allows, in this context, the statistical investigation of spiking neural
networks as well as the mean field analysis of rate encoding neural
populations [3].
We implement a slightly modified version of the Tsodyks-Markram model
for cyclic networks characterized by Mexican-hat type connectivity
profiles for the synaptic weights. The system shows a surprisingly rich set
of dynamical states, even for rings of only four neurons (or neural
populations), such as transient state dynamics [4]. In this case one
observes extended plateaus in the time series of the population firing
rates, indicating well defined transient states (see left panel of Figure 1).
Going beyond the simple mean-field analysis, we present an in-depth
study of the bifurcation diagram, as a function of several bifurcation
parameters, such as the strength I of the external current or the
integration time 1/ of the neural activity. Four classes of distinct limit
cycles are found (see right panel of Figure 1), in addition to stable
fixpoints and saddles. We note, that our results are not only important for
an in-depth understanding of the network effects of working memory,
but also indicate the possibility to construct central pattern generators
using short-term synaptic plasticity.
Acknowledgements: The work of BS was supported by the European
Union and the State of Hungary, co-financed by the European Social
Fund in the framework of TMOP 4.2.4.A/2-11-1-2012-0001 National
Excellence Program.
References
1. Mongillo G, Barak O, Tsodyks M: Synaptic theory of working memory.
Science 2008, 319(5869):1543-1546.
2. Tsodyks MV, Markram H: The neural code between neocortical pyramidal
neurons depends on neurotransmitter release probability. PNAS 1997,
94(2):719-723.
3. Cortes JM, Desroches M, Rodrigues S, Veltz R, Munoz MA, Sejnowski TJ:
Short-term synaptic plasticity in the deterministic Tsodyks-Markram
model leads to unpredictable network dynamics. PNAS 2013,
110(41):16610-16615.
P90
Graph theoretical comparison of functional connectivity between cLTP
treated and untreated microelectrode arrays
Myles Akin1, Rhonda Dzakpasu2, Yixin Guo1*
1
Department of Mathematics, Drexel University, Phildelphia, PA, 19104, USA;
2
Department of Physics, Georgetown University, Washington, DC 20057, USA
E-mail: yixin@math.drexel.edu
BMC Neuroscience 2015, 16(Suppl 1):P90
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 67 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

Figure 1(abstract P89) The firing rates of the respective neural populations in the ring network, and the corresponding phase space speed q
of the trajectory as a function of time (top and bottom plots of the left panel). The bifurcation diagram of the system as a function the external
input I (right panel). The parameter intervals of four possible limit cycles classes denoted by Roman numerals I-IV are indicated by different green/blue
shades. H points denote Hopf-bifurcation, HO is homoclinic bifurcation of saddles, SNC denotes saddle node bifurcation of cycles. The maximal/minimal
amplitude in a limit cycle is denoted with red/green color. Stable (unstable) fixpoints and limit cycles are continuous (dashed) curves

Figure 1(abstract P90) Graph models. (a) cLTP treated MEA network at baseline; (b) cLTP treated MEA network at 5 days past baseline

Analyzing graph properties of neural networks has recently gained much
attention in attempts to understand how information is processed in the
brain. Using in-vitro techniques to form neural networks has increased in
popularity as it allows one to develop small, easy to record networks that
maintain many of the graph properties of larger brain networks [1]. One
widely recognized tool for studying in vitro networks is the
Microelectrode Array (MEAs) on which neurons can be cultured and
recorded simultaneously. MEAs can be used to grow neural networks
from disassociated cells to understand how neurons spontaneously
connect to create networks and how these networks then evolve over
time. In addition, these cultures can be treated with pharmacological
agents to study how these agents affect the networks as a whole [2,3].
To understand the network formation of MEA cultured neurons, we study
the graph theoretical properties of two MEAs networks, the control MEA
network and the MEA network treated with chemical Long Term
Potentiation (cLTP). The data sets for each MEA network consists of
recording from three days: baseline, 2 days past baseline and 5 days past
baseline. Based on these data sets and the assumption that each electrode
on the MEA records one neuron, we construct functional connectivity
graphs of MEA networks for different days. Nodes in such a connectivity
graph represent the electrodes (also neurons). To determine whether there
is a connection (an edge on the graph) between two nodes, we carry out
several steps of computations. We first filter the recorded spike trains with
a Gaussian kernel, and perform cross-correlation analysis using the Pearson
product moment correlation coefficient [4]. We set a correlation threshold
by applying a shuffling method to the inter-spike intervals of a spike train.
Using thresholded correlations, unweighted, undirected adjacency matrices,
we create corresponding graphs for untreated (not shown) and treated MEA
networks (shown baseline and 5 days past baseline in Figure 1). We find that
the synchronization and average node degree increase dramatically for the
cLTP treated networks while the untreated network shows no obvious change.
To better understand the treated and untreated MEA network, we will
evaluate the graph theoretic properties, such as degree distribution and
clustering coefficient. We will determine how cLTP affects these properties.
The graphical analysis will enable us to identify what type of network each
is (such as a small-world or a scale free network) and determine whether
cLTP has an effect on the network development or merely on the strength
of connectivity. We conjecture that cLTP treated networks have more
efficient and quicker communication between nodes. Therefore, the cLTP
treated networks show greater clustering as well as shorter path length than
the untreated networks. Information flow is another important aspect of
such graph model. We intend to develop directed graphs using transfer
entropy to study how information flow of the network may change during
its development.
References
1. Bullmore E, Sporns O: Complex brain networks: graph theoretical analysis
of structural and functional systems. Nature Reviews 2009, 10:186-198.
2. Niedrinhaus M, Chen X, Conant K, Dzakpasu R: Synaptic potentiation
facilitates memory-like attractor dynamics in cultured in vitro
Hippocampal networks. PLOS One 2013, 8(1).
3. Vincent K, Tauskela J, Mealing G, Thivierge JP: Altered network
communication following neuroprotective drug treatment. PLOS One
2013, 8(1).
4. Segev R, Baruchi T, Hulata E, Ben-Jacob E: Hidden Neuronal Correlation in
Cultured Networks. Physical Review Letters 2004, 92:118102-2-118102-4.
P91
Simulation of AMPA and NMDA contribution to postsynaptic response
Crhistian Miguel Gutirrez Galindo1*, Virginia Gonzlez Vlez1, Amparo Gil2
1
Dept. Ciencias Bsicas, Universidad Autnoma Metropolitana Azcapotzalco,
02200, Mxico DF, Mxico; 2Dept. MACC, Universidad de Cantabria, 39005,
Santander, Spain
E-mail: cr_mgg@hotmail.com
BMC Neuroscience 2015, 16(Suppl 1):P91
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P91) Simulations A Low-concentration pulses. B High-concentration pulses. Above input trains (red) and below AMPA (green),
NMDA (magenta) and Total receptors (black) responses
The synaptic response depends on the dynamics of the released
neurotransmitter, as well as on the ability of ionic receptors (AMPA and
NMDA) to activate and deactivate in time. Simulation of these receptors
allow to study interesting patterns as neuronal plasticity, which modulates
the efficiency of transmission in two different manners: facilitation
(incremented response to weak stimuli), and depression (decreased
response to strong stimuli). In this work, we have simulated the
postsynaptic response due to AMPA and NMDA activation. Our model
considers the postsynaptic membrane as a disk of 0.1 microns of radius
where receptors lie with a ratio of 4:1 (80 AMPA and 20 NMDA) [1]. They
are uniformly distributed in the whole area of the disk, assuming they have
the same probability of being activated as they come in contact with
neurotransmitter molecules which are dispersed homogeneously over the
postsynaptic membrane. We use the kinetic models proposed by [2] and
[3] for AMPA and NMDA receptors, respectively. The model was solved
with a stochastic simulation algorithm, and plots show the mean values
calculated from 500 runs. Figures 1A and 1B show the results obtained
when the neurotransmitter comes in the form of two pulses of different
concentrations in a short period of time. Figure 1A corresponds to the case
of low concentration pulses (40 and 160 molecules for the first and second
pulses, respectively). In this simulation, it can be seen how AMPA receptors
increase their response to the second stimulus, exhibiting facilitation.
Figure 1B shows the case of high concentration pulses (1207 and 4787
molecules, respectively). In this simulation, a desensitized response is
observed, that is, a depression effect which reduces the number of open
receptors for the second stimulus. In both cases it is observed that the
contribution made by the NMDA receptors is identical, whereas AMPA
receptors have a contribution that follows the rapid dynamics of the
neurotransmitter released by the presynaptic neuron.
Other simulations for different trains of neurotransmitter were performed,
and similar results were obtained. We also performed simulations for
different stimulation frequencies in larger timescales and again, AMPA
receptors followed rapid neurotransmitter dynamics while NMDA
receptors exhibited changes only in time activation. Then, we conclude
that the contribution of AMPA receptors is the most important one in
short timescales, whereas NMDA receptors participate mainly in larger
timescale responses. These results are the basis of our current work on
proposing a postsynaptic model intended to work with our presynaptic
model [4].
Acknowledgements: C.M.G. thanks CONACYT for the financial support
through his graduate scholarship to carry out this investigation
References
1. Franks MKevin, Bartol MThomas Jr, Sejnowski JTerrence: A Monte Carlo
model reveals independent signaling at central glutamatergic synapses.
Biophysical Journal 2002, 83:2333-2348.
2. Jonas P, Major G, Sakmann B: Quantal components of unitary EPSCs at
the mossy fibre synapse on CA3 pyramidal cells of rat hippocampus. The
Journal of Physiology 1993, 472:615-663.
3. Heckmann M, Bufler J, Franke C, Dudel J: Kinetics of homomeric GluR6
glutamate receptor channels. Biophysical Journal 1996, 71:1743-1750.
Page 68 of 200
4. Gil A, Gonzlez V: Exocytotic dynamics and calcium cooperativity effects
in the calyx of Held synapse: A modeling study. Computational
Neuroscience Journal 2010, 2b8:65-76.
P92
Detecting network states in white noise
Jaroslav Hlinka1*, Michal Hadrava1,2
1
Institute of Computer Science, Czech Academy of Sciences, Prague, Czech
Republic; 2Faculty of Electrical Engineering, Department of Cybernetics,
Czech Technical University in Prague, Prague, Czech Republic
E-mail: hlinka@cs.cas.cz
BMC Neuroscience 2015, 16(Suppl 1):P92
Nonstationarity of neural dynamics is a ubiquitous property that is crucial to
understanding many key phenomena of both healthy and diseased brain
function, including circadian rhythms, dynamics of epileptic activity as well
as cognitive processing. Detecting switching of brain states has recently
become of growing interest in the human brain neuroimaging community.
However, from the data analysis/modelling perspective the task is quite
challenging, and competing approaches exist [1]. One widely adopted
approach is the use of clustering methods in the temporal domain to detect
temporally contiguous clusters of time points with a similar structure of
some instantaneous property - e.g. neural activity or functional connectivity
profile. While this approach may in principle help to explore the switching
structure of brain dynamics, it comes with technical challenges related the
presence of noise in both the dynamics and measurements. In particular, as
we documented in a recent study [2], comparison of the results with an
appropriate null hypothesis is necessary to avoid spurious detection of
nonstationarity markers such as switching of neural network states.
We document this danger by applying an example analysis pipeline used
in [3] to simulated EEG datasets. The simulated data are generated as
realizations of temporally white noise process (either spatially uncorrelated
or spatially correlated in a pattern corresponding to real EEG data). In each
case, one hundred realizations of a 5 seconds long epoch of N = 20
electrodes (each of 2500 time points corresponding to 2ms sampling
rate). A k-means clustering algorithm with k = 2 to 10 is applied to cluster
the instantaneous synchronization likelihood matrix estimates with
parameters as in [3]. The key observation stable across all setting is that
the applied typical network switching analysis pipeline leads to spurious
discovery of a multitude of network states in the stationary process
realizations, with dominant state duration timescales of several tens to
hundreds milliseconds qualitatively similar to the original results reported
in [3]. These results suggest that observations of network switching should
be always cautiously interpreted and tested against appropriate null
models.
Acknowledgements: We thank Martin Brunovsk from Psychiatric Center
Prague for providing sample EEG data. The research was supported by
the Czech Science Foundation projects No. 13-23940S and No. 13-17187S.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
References
1. Hutchison RM, Womelsdorf T, Allen EA, Bandettini PAD, Calhoun V,
Corbetta M, Duyn JH, Glover GH, Gonzalez-Castillo J, et al: Dynamic
functional connectivity: Promise, issues, and interpretations. NeuroImage
2013, 80:360-378.
2. Hlinka J, Hadrava M: On the danger of detecting network states in white
noise. Frontiers in Computational Neuroscience 2015, 9:11.
3. Betzel RF, Abell M, ODonnell BF, Hetrick WP, Sporns O: Synchronization
dynamics and evidence for a repertoire of network states in resting EEG.
Frontiers in Computational Neuroscience 2012, 6(74):1-13, doi:10.3389/
fncom.2012.00074.
P93
Is it right to estimate inter-modular connectivity from local field
potentials?
Xue-Mei Cui1,2, Won Sup Kim2, Dong-Uk Hwang3, Seung Kee Han2*
1
Normal College, Yanbian University, Yanji 133002, China; 2Department of
Physics, Chungbuk National University, Cheongju, Chungbuk 361-763,
Republic of Korea; 3National Institute of Mathematical Sciences, Daejon
305-811, Republic of Korea
E-mail: skhan@chungbuk.ac.kr
BMC Neuroscience 2015, 16(Suppl 1):P93
Human brains with hundreds of billions of neurons are organized in a
hierarchical modular network. There have been many attempts to estimate
inter-modular connectivity utilizing coherent neuronal activities of a huge
number of neurons, such as the electro-encephalogram, the magneto-
encephalogram, and the functional magnetic resonance imaging. Here we
ask a question: Is the inter-modular connectivity estimated from the
modular activities consistent with the inter-modular connectivity that
could be extracted from the network connectivity of individual nodes?
To answer this question, we introduce a method of estimating the inter-
modular connectivity based on the analysis of inverse phase synchronization
[1,2]. For coupled phase oscillators on a hierarchical modular network shown
in the Figure 1(A), the local field potential corresponding to a module is
defined as the mean phase of oscillators belonging to a subset of the
module.
For strong coupling strength, it is shown in Figure 1(B) that the inverse phase
synchronization index grows linearly with the number of links connecting
two modules. This result enables us to estimate the inter-modular
Figure 1(abstract P93) (A) A hierarchical modular network model 256-p-q-r, where p denotes the number of links of one node with nodes of
its lower-level module, q links with nodes of the rest modules in its upper-level module, and r links with nodes of any modules from the rest
of the network. Here we present an example of p = 13, q = 4 and r = 1. (B) Plot of the inverse phase synchronization index versus the number of links
connecting two modules. The linear dependence between the inverse phase synchronization index and the number of links connecting two modules is
shown when the local field potential is taken over m oscillators belonging to a subset of each module
Page 69 of 200
connectivity in various complex systems from the inverse phase
synchronization index of the mesoscopic modular activities.
References
1. Schelter B, Winterhalder M, Dahlhaus R, Kurths J, Timmer J: Partial phase
synchronization for multivariate synchronizing systems,. Phys. Rev. Lett
2006, 96:208103.
2. Kim S, Cui X-M, Yoon CN, Ta HX, Han SK: Estimating network link weights
from inverse phase synchronization indices. Europhys. Lett 2011, 96:20006.
P94
Analyzing adaptive modulation in spinal motor neurons using multi-
objective evolutionary algorithms
Tomasz G Smolinski1*, Joseph Lombardo2, Melissa A Harrington2
1
Department of Computer and Information Sciences, Delaware State
University, Dover, DE 19901, USA; 2Department of Biological Sciences,
Delaware State University, Dover, DE 19901, USA
E-mail: tsmolinski@desu.edu
BMC Neuroscience 2015, 16(Suppl 1):P94
Spinal motoneurons have long been thought to be simply part of a relay
system that provides rapid, stereotyped outputs for muscles on the basis of a
supraspinal plan tuned by sensory inputs, and activity-dependent plasticity
(ADP) has been presumed to be a property of the brain. However, recent
work indicates that ADP occurs in spinal motor neurons during development,
as well as later in life with skills acquisition and maintenance, and in
response to trauma and disease [1]. Understanding how spinal motoneuron
output can be modified by both increased and decreased activity is thus a
fundamental challenge with implications for athletic training, rehabilitation,
and advanced prosthetics. We hypothesize that that alteration in the
function of Kv7.2 channel (which carries the M current) and changes in
axonal initial segment (AIS) properties are the primary mechanisms of
adaptation of spinal motoneurons to prolonged network activation. This
hypothesis is supported by the literature and our experimental data
demonstrating that persistent activation of spinal cord networks decreases
spinal motoneuron output in a manner consistent with the enhancement of
a sub-threshold, non-inactivating potassium conductance. KCNQ/Kv7
channels, which are non-inactivating potassium channels that activate in the
sub-threshold range [2], are expressed at the AIS, nodes of Ranvier, and soma
of spinal motoneurons [3,4], and modulate their excitability [5,6].
To test our hypothesis, we developed a realistic computational model of
spinal motoneuron activity before and after persistent network activation.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
As the starting point, we utilized a reconstructed spinal motoneuron
morphology of neonatal mice [7] together with the detailed specification
of the active and passive somatodendritic and axonal properties derived
from a rodent cortical neuron model [8]. The model parameter values were
adjusted to match our recordings of motoneuron electrophysiological
properties using a multi-objective evolutionary algorithm (MOEA) [9]. The
algorithm matches multiple selection criteria simultaneously (e.g., spike
frequency, shape, adaptation rate, etc.) and generates entire collections of
neuronal models that can be mined for rules describing the phenomena
captured by the models (for instance, co-regulations between ionic
conductances). Furthermore, since the MOEA generates two independent
databases of models (i.e., before and after persistent activation), we are
able to directly compare the phenomena discovered by our data mining
process in each dataset, thus elucidating the mechanisms underlying
plasticity.
Acknowledgements: Support: NIH 5P20GM103653, NIH 1R15HD075207,
NSF EPSCoR 0814251, NSF HRD1242067.
References
1. Wolpaw JR, Tennison AM: Activity-dependent spinal cord plasticity in
health and disease. Annual Review of Neuroscience 2001, 24:807-843.
2. Schroeder BC, et al: KCNQ5, a novel potassium channel broadly
expressed in brain, mediates M-type currents. J Biol Chem 2000,
275(31):24089-95.
3. Dedek K, Waldegger S: Colocalization of KCNQ1/KCNE channel subunits
in the mouse gastrointestinal tract. Pflugers Arch 2001, 442(6):896-902.
4. Devaux JJ, et al: KCNQ2 is a nodal K+ channel. J Neurosci 2004,
24(5):1236-44.
5. Alaburda A, Perrier JF, Hounsgaard J: An M-like outward current regulates
the excitability of spinal motoneurones in the adult turtle. J Physiol 2002,
540(Pt 3):875-81.
6. Rivera-Arconada I, Lopez-Garcia JA: Effects of M-current modulators on
the excitability of immature rat spinal sensory and motor neurones. Eur
J Neurosci 2005, 22(12):3091-8.
7. Li Y, et al: Developmental changes in spinal motoneuron dendrites in
neonatal mice. J Comp Neurol 2005, 483(3):304-17.
8. Kole MH, et al: Action potential generation requires a high sodium
channel density in the axon initial segment. Nat Neurosci 2008,
11(2):178-86.
9. Patel P, et al: Hybridization of multi-objective evolutionary algorithms
and fuzzy control for automated construction, tuning, and analysis of
neuronal models. BMC Neurosci 2013, 14(Suppl 1):P369.
P95
Synaptic input patterns triggering local dendritic spikes in vivo
Lea Goetz*, Martine R Groen, Arnd Roth, Michael Husser
Wolfson Institute for Biomedical Research, University College London,
London, WC1E 6BT, UK
E-mail: lea.goetz.13@ucl.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P95
Theoretical modelling and experiments in vitro have shown that
the computations performed by single neurons critically depend on the
spatiotemporal patterns of synaptic input to the neuron [1,2]. While
recording the spike output of neuronal populations in vivo is now routinely
possible, the spatiotemporal pattern of synaptic inputs to an entire neuron
is still largely inaccessible to experiment. However, recent experiments in
which sensory-evoked local dendritic spikes were observed in vivo have
provided valuable constraints on possible spatiotemporal patterns of
synaptic input [3-5]. For example, the synaptic input to layer 2/3 pyramidal
neurons in mouse primary visual cortex evokes dendritic spikes at high
frequencies during visual stimulation at the preferred orientation, but not
at non-preferred orientations [3]. Here, we use a biophysical model to
explore which spatiotemporal patterns of synaptic inputs can drive the
observed spikes, and in particular the high frequencies in dendritic spike
bursts.
First, we adapted a detailed active compartmental model of a neocortical
layer 2/3 pyramidal neuron [3] to reproduce biophysical properties and
firing statistics observed in vivo [6], such as firing threshold, mean
membrane potentials for UP and DOWN states, and sparse action
potential firing in the presence of Poisson distributed background
synaptic input. The model generates fast dendritic spikes heterogeneous
Page 70 of 200
in amplitude, time course and spatial extent, as observed in in vivo
experiments [3]. The dendritic spikes, together with somatic action
potential firing, are abolished by blocking the NMDA receptor-mediated
conductance, and their frequency is reduced by hyperpolarization via the
dendritic recording site, again as observed in experiments [3].
Next we analyzed which spatiotemporal synaptic input patterns precede
the generation of local dendritic events. We find that local spikes are
preferentially triggered by excitatory synapses which are spatially and
temporally clustered in the local dendritic branch and neighbouring
branches. While there is great variability in the spatial distribution of
synapses triggering local dendritic spikes, temporal patterns of excitatory
synaptic input are stereotyped and tend to be sparse. High-frequency
dendritic spikes similar to those observed in vivo occur preferentially
when excitatory synaptic inputs are interspersed with inhibitory synaptic
inputs. Furthermore, local spikes preceding backpropagating action
potentials lead to the highest instantaneous dendritic spike frequencies.
The highest sustained frequencies are generated by local spikes initiated
sequentially in several neighbouring dendritic branches. Notably, only
some of the high frequency dendritic events are effective in evoking
action potential output, while others remain local. We use cluster analysis
to establish the local and global conditions under which specific
spatiotemporal patterns of synaptic input can influence neuronal output.
In summary, we have derived spatial and temporal rules for synaptic
input and identify input patterns that locally exploit the non-linear
integration capacities of dendrites. Which computations can be
implemented by spiking dendrites receiving such synaptic input patterns?
These constraints on synaptic input patterns, together with the realistic
intrinsic properties displayed by the model, put us in a position to
investigate the role of dendritic excitability in shaping the input-output
relation of the neuron.
References
1. Poirazi P, Brannon T, Mel BW: Pyramidal neuron as two-layer neural
network. Neuron 2003, 37:989-999.
2. Branco T, Clark BA, Husser M: Dendritic discrimination of temporal input
sequences in cortical neurons. Science 2010, 329:1671-1675.
3. Smith SL, Smith IT, Branco T, Husser M: Dendritic spikes enhance
stimulus selectivity in cortical neurons in vivo. Nature 2013, 503:115-120.
4. Sheffield MEJ, Dombeck DA: Calcium transient prevalence across the
dendritic arbour predicts place field properties. Nature 2015, 517:200-204.
5. Palmer LM, Shai AS, Reeve JE, Anderson HL, Paulsen O, Larkum ME: NMDA
spikes enhance action potential generation during sensory input. Nat.
Neurosci 2014, 17:383-390.
6. Waters J, Helmchen F: Background synaptic activity is sparse in
neocortex. J. Neurosci 2006, 26:8267-8277.
P96
Functional consequences of non-equilibrium dynamics caused by
antisymmetric and symmetric learning rules
Dmytro Grytskyy1*, Markus Diesmann1,2,3, Moritz Helias1
1
Institute of Neuroscience and Medicine (INM-6) and Institute for Advanced
Simulation (IAS-6), Jlich Research Centre and JARA, Jlich, Germany;
2
Department of Psychiatry, Psychotherapy and Psychosomatics, Medical
Faculty, RWTH Aachen University, Aachen, Germany; 3Department of Physics,
Faculty 1, RWTH Aachen University, Aachen, Germany
E-mail: d.grytskyy@fz-juelich.de
BMC Neuroscience 2015, 16(Suppl 1):P96
Connectivity in the brain is asymmetric, which is evident by Dales principle
of excitatory and inhibitory neurons. As a consequence, biologically realistic
neuronal networks cannot be in thermodynamic equilibrium [1]. Even in a
stationary state, probability fluxes perpetuate, leading to non-equilibrium
steady states [2,3]. We here investigate the computational consequences of
non-equilibrium dynamics for synaptic plasticity and learning. Formulating
the stochastic dynamics in sparsely connected networks of non-linear
Langevin equations in terms of path integrals [4], we show that biologically
plausible correlation-sensitive plasticity rules follow from first principles. We
investigate two different rules: 1.) Maximizing a measure of irreversibility [2]
by gradient descent with respect to the weights, we obtain a local learning
rule sensitive to the derivative of the covariance between the pre- and
postsynaptic neuron. The obtained rule can be interpreted as spike-timing
dependent plasticity (STDP) [5,6] with a narrow antisymmetric learning
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
window. We show that the learning rule increases synaptic weights in the
direction of the (direct or indirect) causal influence between a pair of
neurons. In this way indirect causal relationships are transformed into
strengthened direct connections. An instability of the state with vanishing
weights manifests itself as spontaneous symmetry breaking and the
divergence of connection weights. 2.) We investigate a rule that is sensitive
to the zero-time-lag covariance. It can be considered as a complement to
the first rule approximating STDP with a symmetric narrow learning window.
This rule results in the strengthening of loops, of mutual connections
between neurons getting common input, and of indirect (also backward)
connections between neurons a and b if a influences b. We derive analytical
expressions that describe these effects quantitatively. We show how
nonlinearities in the neuronal transmission naturally stabilize synaptic
weights and how the limited dynamic range of neuronal activity mediates
competition between synapses. The work elucidates the tight connection
between measures of reversible and irreversible dynamics and the
structures resulting from learning rules that optimize either of the two.
Acknowledgements: Partially supported by the Helmholtz portfolio
theme SMHB, the Jlich Aachen Research Alliance (JARA), HGF Young
Investigators Group VH-NG-1028, EU Grants 269921 (BrainScaleS) and
604102 (HBP).
References
1. Sompolinsky H: Statistical mechanics of neuronal networks. Physics Today
1988, 70-80.
2. Seifert U: Stochastic thermodynamics, fluctuation theorems, and
molecular machines. Rep Prog Phys 2012, 75:126001, doi:10.1088/0034-
4885/75/12/126001.
3. Evans DJ, Morriss GP: Statistical Mechanics of Non Equilibrium Liquids
London: Academic Press 1990.
4. Wio HS, Colet P, San Miguel M: Path-integral formulation for stochastic
processes driven by colored noise. Phys Rev A 1989, 40:7312-7324.
5. Bi GQ, Poo MM: Synaptic modifications in cultured Hippocampal
neurons: dependence on spike timing, synaptic strength, and
postsynaptic cell type. J Neurosci 1998, 18:10464-72.
6. Morrison A, Diesmann M, Gerstner W: Phenomenological models of
synaptic plasticity based on spike timing. Biol Cybern 2008, 98:459-478.
P97
Identifying and exploiting the anatomical origin of population rate
oscillations in multi-layered spiking networks
Hannah Bos1*, Jannis Schuecker1, Markus Diesmann1,2,3, Moritz Helias1
1 BMC Neuroscience 2015, 16(Suppl 1):P98
Institute of Neuroscience and Medicine (INM-6) and Institute for Advanced
Simulation (IAS-6), Jlich Research Centre and JARA, Jlich, Germany;
2
Department of Psychiatry, Psychotherapy and Psychosomatics, Medical
Faculty, RWTH Aachen University, Aachen, Germany; 3Department of Physics,
Faculty 1, RWTH Aachen University, Aachen, Germany
E-mail: h.bos@fz-juelich.de
BMC Neuroscience 2015, 16(Suppl 1):P97
Fast oscillations of the population firing rate in the high gamma range
(50-200 Hz), where individual neurons fire slowly and irregularly, are
observed in the living brain and in network models of leaky integrate-
and-fire (LIF) neurons, that have also been studied analytically [1].
However, a systematic approach identifying sub-circuits responsible for
specific oscillations in a structured network of neural populations is
currently unavailable.
We consider a large-scale, neural network consisting of 4 layers each
composed of an excitatory and inhibitory population of LIF-neurons with
connectivity determined by electrophysiological and anatomical studies
[2]. In simulations we observe a peak in the power spectrum around
83 Hz in all populations and low frequency oscillations with smaller
power in a subset of the populations. Mapping the dynamics of the
fluctuations to an effective linear rate model, using the recently derived
transfer function for LIF-neurons with synaptic filtering [3], we derive the
spectra of the population firing rates analytically.
Decomposing the noise-driven fluctuations into eigenmodes of the
effective connectivity, we identify the modes responsible for peaks in the
spectra. Applying perturbation theory, we quantify the influence of
individual anatomical connections on the spectrum at given frequencies
and identify a sub-circuitry, localized in the supra-granular and granular
layer, generating the oscillation. These findings are in agreement with
Page 71 of 200
layer-specific local field potential measurements in the Macaque primary
visual cortex, where gamma-frequency oscillations were mostly
pronounced in layer 2,3 and 4B [4]. We exploit this method i) to identify
the connectivity loops responsible for the observed peaks and ii) to alter
the circuitry in a targeted manner to control the position and amplitude
of the peaks and the generation of slow frequency fluctuations. This
requires removal and addition of only small numbers of synapses. The
analytical framework moreover explains the suppression of higher
frequencies by distributed delays and the amplification of population
specific oscillatory input. Mapping the stimulus vector onto the
eigenmodes of the system shows how the components of the input
vector are processed in the network. Thus one can derive the sensitivity
of the population rate dynamics to the direction and frequency of stimuli.
Our method finds application in the identification of the connectivity
loops that determine emergent and externally driven global measures of
activity observable in experiments as well as in engineering circuits that
exhibit desired correlations on the population level.
Acknowledgements: We acknowledge funding by the Helmholtz
Association: portfolio theme SMHB and Young Investigators Group VH-NG-
1028, and EU Grants 269921 (BrainScaleS) and 604102 (Human Brain
Project). All network simulations were carried out with NEST (http://www.
nest-initiative.org).
References
1. Brunel N, Hakim V: Fast Global Oscillations in Networks of Integrate-and-
Fire Neurons with Low Firing Rate. Neural Comput 1998, 1621-1671.
2. Potjans TC, Diesmann M: The cell-type specific cortical microcircuit:
relating structure and activity in a full-scale spiking network model.
Cereb Cortex 2014, 24:785-806.
3. Schuecker J, Diesmann M, Helias M: Spectral properties of excitable systems
subject to colored noise , arXiv:1411.0432 [cond-mat.stat-mech].
4. Xing D, Yeh C-I, Burns S, Shapley MS: Laminar analysis of visually evoked
activity in the primary visual cortex. PNAS 2012, 109(34):13871-13876.
P98
Homeostatic intrinsic plasticity, neural heterogeneity and memory
maintenance
Yann Sweeney1,2*, Jeanette Hellgren-Kotaleski2, Matthias Hennig1
1
IANC, School of Informatics, University of Edinburgh, Edinburgh, UK;
2
Department of Computational Biology, KTH, Stockholm, Sweden
E-mail: yann.sweeney@ed.ac.uk
Neural firing rates must be maintained within a stable range in the face of
ongoing fluctuations in synaptic activity. This can be achieved through
homeostatic intrinsic plasticity. However, here we show that such a
mechanism, while successfully regulating neural firing rates, has an
adverse effect on a networks ability to encode and retain memories. This
is due to its interactions with Hebbian plasticity; neurons whose firing rates
change following potentiation or depression of synaptic inputs will
experience modifications in intrinsic excitability toward their homeostatic
target, which can cause subsequent synaptic weight variations and disrupt
learning. Essentially, this failure is a direct consequence of homeostasis
preventing neural heterogeneity in order to maintain stable activity.
We propose a new mechanism, diffusive homeostasis, in which neural
excitability is modulated by a diffuse messenger, specifically nitric oxide,
which is known to freely cross cell membranes and homeostatically
regulate neural excitability [1]. Information about a neurons firing rate can
be carried by nitric oxide, meaning that an individual neurons excitability
is affected by neighbouring neurons firing rates as well as its own. We find
that this allows a neuron to deviate from the target population activity, as
its neighbours will counteract this deviation, thus maintaining stable
average activity. We show that this form of neural heterogeneity endows a
network with more flexibility than heterogeneity through variable target
firing rates in individual neurons, which in turn leads to networks that are
more responsiveness to changes in synaptic inputs (Figure 1) [2]. The
increased flexibility in firing rates conferred by diffusive homeostasis
resolves the conflict between homeostatic intrinsic plasticity and Hebbian
plasticity by limiting the impact of homeostasis on individual synaptic
modifications. Consequently, networks endowed with this diffusive
mechanism have an improved learning capability compared to canonical,
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P98) Responsiveness of a network to changes in input when different types of homeostasis are used in order to reach a
steady target state. Panels A-B, D-E show firing rate changes of all neurons in a network following input changes . Black lines show linear fit, with
R2 values used to quantify the average network responsiveness (panel C)
local homeostatic mechanisms, exhibit more stable synaptic weights, and
allow for more efficient use of neural resources.
This work was supported by the Erasmus Mundus EuroSPIN programme
(YS) and MRC Fellowship G0900425 (MHH).
References
1. Steinert JR, Robinson SW, Tong H, Haustein MD, Kopp-Scheinpflug C,
Forsythe ID: Nitric oxide is an activity-dependent regulator of target
neuron intrinsic excitability. Neuron 2011, 71(2):291-305.
2. Sweeney Y, Hellgren-Kotaleski J, Hennig M: A diffusive homeostatic signal
maintains neural heterogeneity and responsiveness in cortical networks.
biorxiv 2015 [http://dx.doi.org/10.1101/011957].
P99
Brain state dependent stimulus information in the auditory
thalamocortical system
Jon Bamber1*, Shuzo Sakata2, J Michael Herrmann3
1
Institute for Adaptive and Neural Computation, University of Edinburgh,
Edinburgh, EH8 9AB, UK; 2Strathclyde Institute of Pharmacy and Biomedical
Sciences, University of Strathclyde, Glasgow, G4 0RE, UK; 3Institute of
Perception, Action and Behaviour, University of Edinburgh, Edinburgh, EH8
9AB, UK
E-mail: jon.bamber@ed.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P99
Activity in the absence of stimuli is ubiquitous across the thalamocortical
system (TS), with patterns of spontaneous activity reflecting ongoing
behavioural state. Under anaesthesia and during deep sleep the TS
operates in an inactivated state (characterised by low frequency high
amplitude oscillations in local field potential (LFP)) in which neurons
collectively alternate between periods of local silence and high synaptic
activity. During wakefulness and REM sleep, however, the TS operates in an
activated state (characterised by high frequency low amplitude oscillations
in LFP) in which neurons fire in a sustained desynchronised manner. Such
brain states may be indicative of different modes of neural processing, but
the effect of brain state on neural processing remains unclear.
Here we analyse data recorded in the auditory TS of urethane anaesthetised
rats subjected to single-click acoustic stimulation over a range of intensities,
presented in both the inactivated state (natural under the anaesthesia) and
the activated state (induced through electrical stimulation of the basal
forebrain). Evoked spike trains were identified for single units in the auditory
thalamus and across depths of the primary auditory cortex. Mutual
Page 72 of 200
information (MI) between stimulus and response was then computed for
spike probability, counts and timing.
Evoked spiking activity was observed to last around 300ms, consisting of
distinct initial and secondary (rebound) activity. Analysis of responses of
single units over a full 300ms window showed that spike count and timing
measures gave little more MI than response probability, suggesting that
stimulus intensity is primarily encoded probabilistically, at least at the level
of the single unit. Moreover, whilst many single units are uninformative of
stimulus intensity, the most informative thalamorecipient (TC) and
infragranular (IF) single units show increased MI in the activated state.
Additionally, upon temporally partitioning data according to initial or
rebound activity, we see that TC units are more informative in the
activated state during initial activity whereas IF units are more informative
in the activated state during rebound activity, suggesting that information
loss in the inactivated state may be cumulative in time as sensory signals
propagate through neural circuits.
P100
Different roles for ipsilateral positive feedback and commissural
inhibitory networks in oculomotor velocity to position neural
integration
Keiichiro Inagaki*, Yutaka Hirata*
Department Robotic Science and Technology, Chubu University, Kasugai,
Aichi, 487-8501, Japan
E-mail: Inagakikay@isc.chubu.ac.jp; yutaka@isc.chubu.ac.jp
BMC Neuroscience 2015, 16(Suppl 1):P100
Saccadic eye movements are made about twice a second to shift our gaze
either consciously or unconsciously. Brainstem burst neurons trigger each
saccade with a bursting activity closely related to saccadic velocity that
must be transformed into persistent firing activity to maintain post-
saccadic eye position. A conceptual neural mechanism achieving this
velocity to position integration is the oculomotor neural integrator (NI).
Neurons in the nucleus of prepositus hypoglossi (NPH) in mammals and
equivalently Area I (AI) in goldfish have been demonstrated to exhibit
persistent activity relating to eye position. Ipsilateral positive feedback
found in NPH and AI [1], and commissural inhibition between these
bilateral areas [2] are proposed to be the two candidate neuronal networks
for the integration. However, how these networks contribute to generate
persistent firing is still unknown. Recent behavioral and optogenetic
experiments demonstrated that eye position holding properties are
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
individually modifiable for nasal and temporal hemi-fields of the eye,
thereby suggesting existence of at least two NIs in each nucleus [3,4]. In
the present study, we constructed a NI neuronal network model closely
following anatomical and physiological evidence to better understand the
roles of the two NI networks.
The model consists of an excitatory burst neuron (EBN), an inhibitory
burst neuron (IBN), a tonic neuron (TN), and a bilateral NI network in
which 15 integrator neurons (INs) are included unilaterally. Each IN
receives input from TN, EBN, and IBN, and also has positive recurrent
feedback connections from all ipsilateral INs including commissural
inhibitory connections from all contralateral INs. The output of EBN, IBN,
and each NI are weighted-summed at a motor neuron whose output is
sent to an eye plant model to simulate eye movements. These neuron
models are described as conductance-based spiking neurons. Synaptic
weights in the model were carefully tuned to reproduce saccades and
post-saccadic stable eye positions. Following individual modification of
either the synaptic weights of ipsilateral feedback connection or those of
commissural inhibition we measured post-saccadic eye position drift.
Simulation results showed that persistent firing activity of INs is maintained
principally by the ipsilateral positive feedback. By contrast, commissural
inhibition contributed mainly to null eye position which is the asymptote
from which the NI becomes leaky. These findings are the first
demonstration by a NI neuronal network model assigning specific and
different roles to the ipsilateral positive feedback and the commissural
inhibition. In future work, we will use this model to evaluate the elective
eye holding properties within the two hemi-field NIs.
Acknowledgement: We thank Prof. Robert Baker for his valuable
comments on this research. This work was supported by JSPS KAKENHI
Grant Number 24300115 (to Y. Hirata).
References
1. Seung HS, Lee DD, Reis BY, Tank DW: Stability of the memory of eye
position in a recurrent network of conductance-based model neurons.
Neuron 2000, 26(1):259-271.
2. Aksay E, Olasagasti I, Mensh BD, Baker R, Goldman MS, Tank DW:
Functional dissection of circuitry in a neural integrator. Nature
Neuroscience 2007, 10(4):494-504.
3. Okamura N, Baker R, Hirata Y: Monocular eye position specificity in the
oculomotor neural integrator. BMC Neuroscience 2011, 12(Suppl 1):151.
4. Goncalves PJ, Arrenberg AB, Hablitel B, Bair H, Machens CK: Optogenetic
perturbations reveal the dynamics of an oculomotor integrator. Frontiers
in Neural Circuits 2014, 8(10):1-21.
P101
A categorical approach to neurodynamical modelling of musical
tonality
Michal Hadrava1,2*, Jaroslav Hlinka2
1 2
Department of Cybernetics, Faculty of Electrical Engineering, Czech
Technical University in Prague, Prague, 166 27, Czech Republic; 2Department
of Nonlinear Dynamics and Complex Systems, Institute of Computer Science,
The Czech Academy of Sciences, Prague, 182 07, Czech Republic
E-mail: hadrava@cs.cas.cz
BMC Neuroscience 2015, 16(Suppl 1):P101
Music gives rise to some of the strongest emotional experiences in our
lives. Both musicological and music-psychological evidence converge on
the theory that music achieves this by a sophisticated play with our
expectancies [1]. The extent to which a pitch/chord is expected in the
context of a musical sequence is closely related to its perceived stability
and also to how much it apparently attracts other pitches/chords in the
sequence (see [2] and the references thereof). The phenomenon wherein
more stable pitches attract the less stable ones is called tonality.
Over the eons, virtually every musical culture has created its own set of
musical scales. Generally, each scale imposes a hierarchy of stability on all
pitches sounded in its context [3]. We call such a hierarchy a tonal hierarchy.
In [4] a formula for computing tonal stability of arbitrary pitch in a given
context was proposed based on a network of neural oscillators. Even though
high correlations with tonal hierarchies obtained experimentally
were achieved by tuning the single parameter of the formula, there were
some notable discrepancies between predictions of the formula and the
psychological data. These might be explained by drastic approximations
made on the way from the complete neurodynamical model to the formula
Page 73 of 200
which were necessitated by the complexity of the model. For instance, all
interactions within the network were neglected. Nevertheless, this line of
research clearly demonstrates the potential of neurodynamics to serve as a
basis for derivation of a general theory of tonality applicable even to
contemporary art music which is our main interest. Indeed, the
neurodynamical model from [4] makes no assumptions whatsoever
regarding its inputs. What is currently missing is a model of a suitable form
so that such a theory can be derived from it without drastic approximations.
We aim to fill this gap.
It has become a standard in neuroscience community to approach
neurodynamics from a dynamical systems perspective. Unfortunately, none
of the approaches to analytical treatment of oscillatory dynamical systems
we have reviewed proved suitable for our purposes. Consequently,
inspired by the so-called memory evolutive neural systems [5], we are
currently developing a neurodynamical model based on category theory.
More specifically, we model the neural rhythms present in the auditory
system during stimulation with a tone/chord as a category with objects
representing the rhythms and morphisms representing the resonant
interactions among them. Response to a sequence of stimuli is modelled
using partial functors between such categories. The preliminary results are
quite promising: from a very simple model we managed to derive a
formula which predicts the ordering of pitches according to tonal stability
within a major scale almost perfectly. Moreover, due to its nature, our
model yields an almost mechanistic explanation of tonality. This, in our
opinion, is as important a feature of any model as a good fit to data.
Acknowledgements: This work was supported by the Grant Agency of
the Czech Technical University in Prague, grant No. SGS14/192/OHK3/3T/
13, and the Czech Science Foundation project No. 13-23940S.
References
1. Huron D: Sweet Anticipation: Music and the Psychology of Expectation
Cambridge: The MIT Press 2006.
2. Woolhouse M: Modelling tonal attraction between adjacent musical
elements. J New Music Res 2009, 38(4):357-379.
3. Krumhansl CL: Cognitive Foundations of Musical Pitch New York: Oxford
University Press 1990.
4. Large EW: A dynamical systems approach to musical tonality. Studies in
Computational Intelligence: Nonlinear Dynamics in Human Behavior Berlin
Heidelberg: Springer-Verlag: Huys R, Jirsa VK 2011, 328:193-211.
5. Ehresmann AC, Vanbremeersch J-P: Memory Evolutive Systems: Hierarchy,
emergence, cognition. Studies in Multidisciplinarity Amsterdam: Elsevier
Science 2007, 4.
P103
Balancing the critical period of spiking neurons with attractor-less STDP
Simon M Vogt1*, Ulrich G Hofmann2
1
BrainLinks-BrainTools, University of Freiburg, Freiburg im Brisgau, Germany;
Neuroelectronic Systems, University Medical Center Freiburg, Freiburg im
Brisgau, Germany
E-mail: simon.vogt@blbt.uni-freiburg.de
BMC Neuroscience 2015, 16(Suppl 1):P103
Attractor-based (multiplicative) STDP has been called a biologically more
realistic form of spike timing dependent plasticity than attractor-less
(additive) variants [1-4], as it produces unimodal distributions of
synaptic weights [5,6] when given poisson-distributed random input spike
trains [1,2,4,7-9]. While unimodal weight distributions have been observed
as an outcome in some in vitro experiments [5], the actual biochemical
process that changes synaptic connection strengths has yet to be fully
understood.
Unfortunately, attractor-based STDP has been repeatedly shown to be
computationally less powerful than attractor-less STDP [9] and successful
implementations of attractor-based STDP used attractors that were either
very weak [8] or very close to some minimum weight [2,3], causing a
more additive-like [2,8] behaviour of the plasticity rule. We therefore
examined possible biological interpretations of attractor-less STDP rules
while keeping in mind that any STDP rule is just an abstraction from the
hidden biophysical reality.
We show how unimodal weight distributions can reliably result from
attractor-less STDP when negative synaptic drift is combined with
activity-independent synaptic growth. A bimodal distribution is then only
formed when non-random (polychromous [10]) poisson-distributed inputs
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
are presented to a neuron [11]. In practice, this produces a plasticity rule
that keeps the postsynaptic neuron unselective and responsive to a
broad range of inputs while receiving only random spikes, but quickly
allows the neurons receptive field to become highly selective as soon as
some inputs start repeating a non-random ordering of spikes.
This stabilization procedure preserves STDPs sensitivity to temporally
shifted correlations in input spike data, which in turn gives us several
beneficial features for biologically more realistic and computationally
more powerful [12] implementations of plasticity in spiking neural
networks.
References
1. Rubin J, Lee D, Sompolinsky H: Equilibrium Properties of Temporally
Asymmetric Hebbian Plasticity. Phys Rev Lett 2001, 86(2):364-367,
doi:10.1103/PhysRevLett.86.364.
2. Gtig R, Aharonov R, Rotter S, Sompolinsky H: Learning input correlations
through nonlinear temporally asymmetric Hebbian plasticity. J Neurosci
2003, 23(9):3697-3714.
3. Morrison A, Aertsen A, Diesmann M: Spike-timing-dependent plasticity in
balanced random networks. Neural Comput 2007, 19(6):1437-1467,
doi:10.1162/neco.2007.19.6.1437.
4. van Rossum MCW, Bi GQ, Turrigiano GG: Stable Hebbian learning from
spike timing-dependent plasticity. J Neurosci 2000, 20(23):8812-8821.
5. Turrigiano GG, Leslie KR, Desai NS, Rutherford LC, Nelson SB: Activity-
dependent scaling of quantal amplitude in neocortical neurons. Nature
1998, 391(6670):892-896, doi:10.1038/36103.
6. Turrigiano GG, Nelson SB: Homeostatic plasticity in the developing
nervous system. Nat Rev Neurosci 2004, 5(2):97-107, doi:10.1038/nrn1327.
7. Sjstrm PJ, Turrigiano GG, Nelson SB: Rate, timing, and cooperativity
jointly determine cortical synaptic plasticity. Neuron 2001, 32(6):1149-1164.
8. Gilson M, Burkitt AN, Grayden DB, Thomas Da, van Hemmen JL: Emergence
of network structure due to [STDP] in recurrent neuronal networks V:
self-organization schemes and weight dependence. Biol Cybern 2010,
103(5):365-386.
9. Billings G, van Rossum MCW: Memory retention and spike-timing-
dependent plasticity. J Neurophysiol 2009, 101(6):2775-2788, doi:10.1152/
jn.91007.2008.
10. Izhikevich EM: Polychronization: computation with spikes. Neural Comput
2006, 18(2):245-282.
11. Guyonneau R, VanRullen R, Thorpe SJ: Neurons tune to the earliest spikes
through STDP. Neural Comput 2005, 17(4):859-879, doi:10.1162/
0899766053429390.
12. Olshausen BA, Field DJ: Emergence of simple-cell receptive field
properties by learning a sparse code for natural images. Nature 1996,
381(6583):607-609, doi:10.1038/381607a0.
P104
Spiking neural network configuration designed for switching between
basic forms of movement in a biped robot
Uziel Jaramillo-Avila, Horacio Rostro-Gonzlez*
Department of Electronics, Engineering Division, University of Guanajuato,
Salamanca, Guanajuato, 36885, Mxico
E-mail: hrostrog@ugto.mx
BMC Neuroscience 2015, 16(Suppl 1):P104
Figure 1(abstract P104) A. Neuronal network including all the required positions for the 4 basic movements and the 30 synapses required
between them. B. Positions of the robot for each one of the 11 neurons. C. Biped robot setup
Page 74 of 200
In this work, we present a neural network design based on a simplified
form and highly suitable for hardware implementation of the integrate-
and-fire spiking neuron model for achieving locomotion in a biped robot.
It is well known that bipedal walking is one of the most complex and
common tasks in robots and humanoids and it is also known that this
problem is generally tackled by using one type of movement at a time, e.
g. forward walking [1]. In this regard, we propose to decompose the
movement of a robot with 6 degrees of freedom and to find a series of
movements that allow it to take a step forward, a step backwards and
make a turn to each side. Thus, a series of 11 commands were
successfully achieved, each of them configuring the robot to a specific
position that by using them in different sequences, it is possible to
perform the movements abovementioned, all starting and ending in a
home position. At the same time, this makes easy switching between
them for the movement of the robot in a real environment. Two
hypotheses are implicit in this study: on one hand, the movements listed
above are equivalent to a spasm or muscle reflex in a biological organism
[2]. On the other hand, each one of the commands that makes a
simultaneous movement in several motors of the robot is controlled by a
spiking neuron, which is attractive because it reflects more closely
biological models, and may be more compatible with biomedical
applications such as neuroprostheses [3]. To validate our approach, we
improved a hardware implementation (on a Spartan 6 FPGA board from
Opal Kelly) of the network shown in Figure 1 and tested it successfully on
a real biped robot from Lynxmotion.
Once the network topology is configured for one of the movements, a
corresponding sequence of neurons is activated as follows:
Step forward: A-B-E-H-I-J-U-W.
Step backward: A-B-F-L-K-V-M-X.
Turn left: A-B-E-N-P-J-S-T.
Turn right: C-D-G-Q-O-H-R-T.
Acknowledgements: This work has been partially funded by the CONACYT
Project 229839, the SEP- PRODEP project and the DAIP Project 444 (2014).
References
1. Alexander RMcN: Bipedal Animals, and Their Differences from Humans.
Journal of Anatomy 2004, 204(5):321-330.
2. Geng T, Porr B, Florentinwrgtter B: A reflexive neural network for
dynamic biped walking control. Neural computation 2006, 18(5):1156-1196.
3. Lewis MA, Tenore F, Etienne-Cummings R: CPG design using inhibitory
networks. Robotics and Automation, 2005. ICRA 2005. Proceedings of the
2005 IEEE International Conference on IEEE 2005, 3682-3687, April.
P105
Dynamics analysis of neural univariate time series by recurrence plots
Tamara Toi1*, Peter beim Graben2,3, Kristin K Sellers4, Flavio Frhlich4,
Axel Hutt1
1
Inria, CNRS, Loria, UMR n 7503, Vanduvre-ls-Nancy, F-54500, France;
2
Institute of German Studies and Linguistics, Humboldt-University, 10178,
Berlin, Germany; 3Bernstein Center for Computational Neuroscience, 10178,
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Berlin, Germany; 4University of North Carolina at Chapel Hill, Chapel Hill,
27599, NC, USA
E-mail: tamara.tosic@inria.fr
BMC Neuroscience 2015, 16(Suppl 1):P105
Transients in non-linear biological signals (e.g., population dynamics or
physiological signals) encode an intrinsic behaviour of system dynamics.
We study the problem of detecting dynamical transients given a set of
signal trials. In general case, different biological signals emerge from
different origins and hence exhibit distinct properties that are hard to
grasp. For example, to analyze sleep recordings, one considers rhythms of
the brain, the cardio-vascular and the respiratory systems. The
synchronous analysis of the corresponding time series is an unsolved
problem and extracting information from such signals and their trial
statistics is challenging. In addition, measurement noise and time jitters
between trials may corrupt signals. To attack this demanding problem, we
start by a preliminary study of extracting features in multiple trials from
univariate time series of the same origin, but without taking into account
the common origin. The new method jointly analyzes neural signals by
extracting statistical properties, obtained by exploiting the fundamental
feature of dynamical systems, the recurrence structure.
Recurrences represent instances in time when the systems trajectory
returns to a phase space neighborhood of a location it has visited
previously, cf. [1]. Classical recurrence plots (RP) [2] represent symmetric
binary valued matrices and characterise the systems phase space. Since
neural recurrence patterns occur in particular frequency bands, we
propose to build novel, frequency-selective recurrence plots from their
time-frequency signal representations. To evaluate the recurrence results,
Figure 1(abstract P105) Detection of dynamical transients. (A) Superposition of several transient oscillations of different frequencies. (B) experimental
Local Field Potentials measured in ferret visual cortex in the presence of a periodic 1Hz visual stimulus [3] starting at sample 0 with a period of 100
samples. Time series are given in left column. Statistically significant data (white values) of recurrence points are illustrated in right column figures (Chi-
square test, p = 0.1)
Page 75 of 200
we compute surrogate RPs by time randomisation of the original time-
frequency representations. To obtain statistically significant recurrence
information, statistics of the original RPs is pixel-wise compared with the
surrogate dataset statistics. This novel inference test provides statistically
significant pixel values and the corresponding recurrence structure.
Results in Figure 1 show an artificial and neural time series and
corresponding statistically significant recurrence plots. One observes that
the method extracts well several transient oscillations of different
frequencies. The analysis of the ferrets data reveals that the first
stimulation response at t = 0ms is recurrent to the second response one
stimulation period later, while the third response at t = 3s is different
from the first two. This result indicates temporal neural adaptation during
the stimulation.
Acknowledgements: Research reported in this publication was supported
by (1) the LIRA, a research initiative of Fraunhofer Society, Philips and Inria
(2) the National Institute Of Mental Health of the National Institutes of
Health under Award Number R01MH101547. The content is solely the
responsibility of the authors and does not necessarily represent the official
views of the National Institutes of Health.
References
1. beim Graben P, Hutt A: Detecting recurrence domains of dynamical
systems by symbolic dynamics. Physical Review Letters 2013,
110(15):154101.
2. Marvan N, Romano MC, Thiel M, Kurths J: Recurrence plots for analysis of
complex systems. Physiscs reports 2007, 438:237-329.
3. Sellers KK, Bennett DV, Frohlich F: Frequency-Band Signatures of Visual
Responses to Naturalistic Input in Ferret Primary Visual Cortex during
Free Viewing. Brain Research 2015, 1598:31-45.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
P106
Simulating electrode arrangements on microelectrode arrays
Inkeri Vornanen*, Kerstin Lenk, Jari AK Hyttinen
Tampere University of Technology, Department of Electronics and
Communications Engineering, BioMediTech, Tampere, Finland
E-mail: inkeri.vornanen@tut.fi
BMC Neuroscience 2015, 16(Suppl 1):P106
Neuronal networks are often studied in vitro using micro-electrode arrays
(MEAs), where neurons are cultured on top of an electrode grid, and the
action potentials can be recorded. This way the electrical activity of the
network can be inspected at multiple locations simultaneously, which
enables the studying of network behavior. A typical MEA has 60 of
electrodes located 50-200 microns between electrodes. However, the
neuronal network has consists of thousands of neurons, so only small
sample of the neurons in the network are recorded. In this study, we
inspected how well different typical electrode arrangements can capture
the network behavior. Therefore we simulated neuronal networks, where
the action potentials were recorded with different electrode arrangements.
We simulated the network using the INEX model [1], which consists of
spontaneously active excitatory and inhibitory neurons. 1005 neurons
were positioned in a grid inside a circle with a 1mm radius and
connected to ~100 nearest neighbors. Different subsets of neurons were
chosen for analysis (see Figure 1) modelling various MEA ensembles:
every 1-10th neuron (panels A-J), the outer- and inner most neurons
(K-L), and different sized grid formations: 3 3 = 9 electrodes (M-V), 8
8 = 64 electrodes (W-Y) and 16 16 = 256 electrodes (Z). Thus panel A
represents the entire network. We calculated the spike and burst rates for
the selected neurons, and compared these between the different sets of
recorded neurons. The bursts were detected using the CMA algorithm [2].
The spiking and bursting rates of neurons in different arrangements are
shown in the Figure 1. In these simulations the neurons on the edges
spike and burst less than the neurons in the middle (compare panels K
and L), due to different neighborhoods. This resembles biological
networks, where parts of the network can be more active than other.
Typically, a lower number of recorded neurons results in low variability of
spike rates (e.g., panels A-J), which in some cases results in erroneous
median values (e.g., panel G) compared to panel A showing the activity
of the whole network. Also when the recorded neurons cover the entire
area of network, the recorded neurons represent better the behavior of
the network, thus even low number of electrodes provide (3x3 grid (M-V))
sufficient results.
Figure 1(abstract P106) The spiking and bursting rates of the neurons in different arrangements in panels A-Z. The red and blue colors correspond to
two networks. The middle black line is the median, the box shows the quartiles and the whiskers 1.5x quartiles of the rates
Page 76 of 200
Acknowledgements: This research has been supported by the 3DNeuroN
project in the European Unions Seventh Framework Programme, Future
and Emerging Technologies, grant agreement n296590.
References
1. Lenk K: A simple phenomenological neuronal model with inhibitory and
excitatory synapses. Advances in Nonlinear Speech Processing 2011, 232-238.
2. Kapucu FE, Tanskanen JMA, Mikkonen JE, Yl-Outinen L, Narkilahti S,
Hyttinen JAK: Burst analysis tool for developing neuronal networks
exhibiting highly varying action potential dynamics. Front Comput
Neurosci 2012, 6.
P107
Modeling the interplay between structural plasticity and spike-timing-
dependent plasticity
Richard M George1,2*, Peter U Diehl1,2, Matthew Cook1,2, Christian Mayr1,2,
Giacomo Indiveri1,2
1
Institute of Neuroinformatics, University Zurich, Zurich, Switzerland; 2ETH
Zurich, Zurich, Switzerland
E-mail: rgeorge@ini.uzh.ch
BMC Neuroscience 2015, 16(Suppl 1):P107
Structural Plasticity describes a form of long-term plasticity, in which the
pruning and the creation of synapses lead to the formation of memories in
the topology of a network of neurons. In contrast, classical learning rules
such as spike-timing dependent plasticity (STDP) focus on changing the
efficacy of synapses, for example by looking at the correlation of pre-and
post-synaptic activity in spiking neural networks. Typically, prolonged
correlated activity leads to a long-term potentiation of the synaptic weight,
while anti-correlated activity depresses the weight.
We propose a computational model that combines classical learning rules
with structural changes in spiking neural network architectures that are
based on observations on the morphological changes real biological
synapses undergo during their live-cycle. Our model is based on the
assumption that newly formed synapses are initially silent, due to their lack
of AMPA receptors. In these synapses, only co-activation with other
synapses can lead to postsynaptic potentials, and if this co-activation is not
present for a critical period, the synapse degenerates again [1]. To study
the interaction of structural plasticity and classical STDP learning rules, we
simulated a highly recurrently connected spiking neural network and
presented topological inputs to its neurons. We implemented the triplet
STDP learning rule proposed by Pfister and Gerstner [2], and applied a
structural plasticity rule where a critical period is opened whenever a
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 77 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

Figure 1(abstract P107) Connectivity matrix from input to target neuron population with and without structural plasticity. Top: Randomly
initialized connectivity, learned using triplet STDP. Bottom: Connectivity learned using structural plasticity and triplet-STDP

synaptic weight is decreased below a certain threshold. If the weight does
not manage to reach a set threshold by the end of the critical period, the
synapse is pruned, and a new synapse is instantiated within the network;
otherwise the synapse is maintained. This approach implies a homeostasis
in the number of consolidated synapses in the network, while keeping the
connectivity at a desired level of sparseness. We show in Figure 1
simulation results in which the input topology of the network is first
learned using only STDP, and then, after activating structural plasticity, the
structure of the connectivity matrix itself is adapted such that it reflects
the input topology.
A major advantage of structural plasticity in artificial neural networks is
given by the fact that it allows a drastic increase in performance given a
finite number of synaptic resources. In addition to offering a promising
approach for optimizing performance in software simulated networks, the
model we propose optimizes the usage of resources in dedicated
hardware neural network implementations that are faced with limited
resources for emulating or simulating synaptic connections. This is
particularly relevant for electronic implementations of spiking neural
networks, ranging from GPU-based systems to mixed signal analog-digital
neuromorphic VLSI devices.
Acknowledgements: This work is supported by European Union Seventh
Framework Program (FP7/2007-2013) under grant agreement no.612058
(RAMP) and the SNF grant 200021-143337.
References
1. Poirazi P, Mel BW: Impact of Active Dendrites and Structural Plasticity on
the Memory Capacity of Neural Tissue. Neuron 2001, 29:779-796.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
2. Pfister JP, Gerstner W: Triplets of spikes in a model of spike timing-
dependent plasticity. Journal of Neuroscience 2006, 26(38):9673-9682.
P108
Understanding short-timescale neuronal firing sequences via bias
matrices
Zachary J Roth1*, Yingxue Wang2, Eva Pastalkova2, Vladimir Itskov3
1 properties of neuronal firing sequences. The bias matrix of a given
Department of Mathematics, University of Nebraska-Lincoln, Lincoln, NE
68588, USA; 2Janelia Research Campus, HHMI, Ashburn, VA 20147, USA;
3 of-mass average and captures more complex relationships among
Department of Mathematics, The Pennsylvania State University, University
Park, PA 16802, USA
E-mail: s-zroth1@math.unl.edu
BMC Neuroscience 2015, 16(Suppl 1):P108
The brain generates persistent neuronal firing sequences across varying
timescales. The short-timescale (~100ms) sequences are believed to be
crucial in the formation and transfer of memories. Large-amplitude local
field potentials known as sharp-wave ripples (SWRs) occur irregularly in
hippocampus when an animal has minimal interaction with its environment,
such as during resting, immobility, or slow-wave sleep. SWRs have been
long hypothesized to play a critical role in transferring memories from the
hippocampus to the neocortex [1]. While sequential firing during SWRs is
known to be biased by the previous experiences of the animal, the exact
relationship of the short-timescale sequences during SWRs and longer-
timescale sequences during spatial and non-spatial behaviors is still poorly
understood. One hypothesis is that the sequences during SWRs are replays
Figure 1(abstract P108) Neuronal spike trains (top) are converted
to bias matrices (middle) by computing the probability of pairs of
neurons spiking in a particular order. The correlation between bias
matrices (bottom) is then computed via the angle between the bias
matrices
Page 78 of 200
or preplays of master sequences, which are sequences that closely mimic
the order of place fields on a linear track [2,3]. Rather than particular hard-
coded master sequences, an alternative explanation of the observed
correlations is that similar sequences arise naturally from the intrinsic biases
of firing between pairs of cells. To distinguish these and other possibilities,
one needs mathematical tools beyond the center-of-mass sequences and
Spearmans rank-correlation coefficient that are currently used.
We introduce a new mathematical tool that captures the intrinsic
sequence (Figure 1) contains more detailed information than the center-
different neuronal sequences. This tool enabled us to directly
investigate the relationships among firing sequences across different
conditions: short-timescale sequences (during SWRs) and long-timescale
behavioral sequences (during spatial navigation and wheel running). We
also performed a pharmacological manipulation that resulted in
elimination of theta oscillation (as previously reported in [4]) and
increased the frequency of SWRs. We have found that the pairwise
biases of sequences during SWRs are highly correlated with sequences
during most of the conditions. Moreover, while sequences of neuronal
activations are uncorrelated across different behaviors, the bias matrices
of SWR sequences are highly correlated with those of various behavior
sequences. Our findings provide a new tool for understanding the
structure of short-timescale neuronal sequences and suggest that
intrinsic pairwise biases are likely the underlying mechanism for the
replay/preplay of longer-timescale sequences observed in the
hippocampus [2,3].
References
1. Girardeau G, Benchenane K, Wiener SI, Buzski G, Zugaro MB: Selective
suppression of hippocampal ripples impairs spatial memory. Nat Neurosci
2009, 12(10):1222-1223.
2. Diba K, Buzski G: Forward and reverse hippocampal place-cell
sequences during ripples. Nat Neurosci 2007, 10(10):1241-1242.
3. Dragoi G, Tonegawa S: Preplay of future place cell sequences by
hippocampal cellular assemblies. Nature 2011, 469(7330):397-401.
4. Wang Y, Romani S, Lustig B, Leonardo A, Pastalkova E: Theta sequences
are essential for internally generated hippocampal firing fields. Nat
Neurosci 2015, 18(2):282-288.
P109
Effects of multimodal distribution of delays in brain network dynamics
Spase Petkoski1,2*, Andreas Spiegler1, Timothe Proix1, Viktor Jirsa1,3
1
Aix-Marseille Universit, Inserm, INS UMR_S 1106, 13005, Marseille, France;
2
Aix-Marseille Universit, CNRS, ISM UMR 7287, 13288, Marseille, France;
3
Centre National de la Recherche Scientifique, Marseille, France
E-mail: spase.petkoski@univ-amu.fr
BMC Neuroscience 2015, 16(Suppl 1):P109
Large-scale modeling of the brain is defined by the local oscillatory
dynamics that are superimposed on an architecture based on a
comprehensive map of neural connections in the brain - connectome [1].
Besides coupling strengths, time-delays due to transmissions via tracts are
crucial features of a connectome. They represent a proxy of the spatial
structure (the tract lengths) to the temporal dynamics. Thus, the most
straightforward approach to model brain dynamics in space and time is to
concatenate oscillatory nodes to a connectome-based network.
The analysis that we performed on the experimentally derived
connectome suggests that the tract lengths - distances between different
brain nodes, thus the time delays, follow a multimodal distribution.
Here, we investigated the conceptual implementation of multimodal
distributions of discrete time delays of network links, and its effects on the
mean-field dynamics. Because of the analytical tractability, the Kuramoto
oscillator describes the temporal dynamics of each node, and the links
between the nodes are symmetric but heterogeneous. Hence, we analyze
synchronization in populations of phase oscillators [2], which have the
same distribution of natural frequencies and coupling strengths, but their
structure is defined solely by their different intra- and inter-population
delays.
Assuming a same overall distribution of time delays, several cases are
investigated: from fully random distribution, to two delays-imposed
structures of subpopulations, Figure 1.
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 79 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

Figure 1(abstract P109) Schematic representation of the delay-imposed structure of population of oscillators: with different inter and same
intra delays in A; same inter and intra delays in B; and random distribution of the delays, in C

For all scenarios, mean-field dynamics are analytically obtained [3] and non-stationary with different time-varying frequencies of synchronization
numerically confirmed. Moreover, boundaries and stabilities of different and order parameters for the clusters.
low-dimensional solutions are also investigated. These reveal a split of In summary, the large-scale spatial organization of the brain is integrated
phase dynamics in different clusters, which can be phase shifted, or even in a network model. Using this model, we present the effects of the
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 80 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

multimodal distribution of time delays and the structure that they impose
on the network dynamics such as synchronization. Hence, we stress the role
of the spatial organization of the brain that is reflected through the different
time-delays between different parts of the brain in the formation of
spatiotemporal dynamics.
References
1. Sporns O, Tononi G, Ktter R: The Human Connectome: A Structural
Description of the Human Brain. PLOS Comput Biol 2005, 1(4):e42,
0010042.
2. Breakspear M, Heitmann S, Daffertshofer A: Generative models of cortical
oscillations: neurobiological implications of the Kuramoto mode. Front
Human Neurosci 2010, 4:190.
3. Ott E, Antonsen TM: Low dimensional behavior of large systems of
globally coupled oscillators. Chaos 2008, 18:037113.

P110
Using the connectome to predict epileptic seizure propagation in the
human brain
Timothe Proix*, Viktor K Jirsa
Aix Marseille Universit, Inserm, INS UMR_S 1106, 13005, Marseille, France
E-mail: timothee.proix@etu.univ-amu.fr
BMC Neuroscience 2015, 16(Suppl 1):P110

Partial seizures in epileptic patients are generated in localized networks,

so-called Epileptogenic Zone (EZ), before recruiting other regions, so-called
Propagation Zone (PZ) [1]. For drug-resistant patients, surgical resection is
sometimes possible. Correctly delineating the extent of the EZ and PZ is
critical for a successful surgical resection, in order to remove enough of the
epileptogenic tissue to prevent seizures while minimizing the cognitive
collateral damages. EZ and PZ extents are evaluated using imaging tools
such as M/EEG, MRI, PET and stereotaxic EEG (sEEG). In this modelisation
work, we used the large-scale connectome to build a network of neural
masses in order to reproduce seizure propagation through the human
brain. In particular, we aimed at predicting the propagation of epileptic
seizures, i.e. the PZ, using the localization of the EZ.
We preprocessed data obtained from 18 different patients with different
types of partial epilepsy. Using MRI and diffusion MRI data, we generated
patient-specific connectomes along with cortical surfaces, using different
parcellation resolutions. Epileptic dynamics of a single region was based on
the Epileptor, a neural mass model able to autonomously generate epileptic
seizures [2]. The different regions interacted via a permittivity coupling
allowing to reproduce seizures propagation such as observed in sEEG [3]. Figure 1(abstract P110) A. Example of the EZ (red) and the
Using a reduced Epileptor model, we performed a stability analysis at the calculated PZ (yellow) displayed on the patient cortical surface,
edge of the seizure onset. We confirmed our results with simulations of the along with sEEG electrodes (small spheres). B. Corresponding
network of Epileptors using The Virtual Brain, a neuroinformatics platform to simulated time series
simulate large-scale dynamics [4]. The analytical prediction of seizure spatial
extent correctly reproduces seizure simulations.
We systematically predicted the spatial spread of the seizure, i.e. the PZ,
for each patient, according to the spatial extent and localization of the EZ 2. Jirsa VK, Stacey W, Quilichini P, Ivanov A, Bernard C: On the nature of
such as observed with sEEG, using different global connectivity and seizure dynamics. Brain 2014, 137:2210-2230.
excitability parameters. An example of an EZ and PZ along with a 3. Proix T, Bartolomei F, Chauvel P, Bernard C, Jirsa VK: Permittivity coupling
simulation of the forward calculation on sEEG electrodes are shown in across brain regions determines seizure recruitment in partial epilepsy.
Figure 1. Our results show a good agreement with clinician predictions, J Neurosci 2014, 34(45):15009-15021.
surgery results, and sEEG signals. To confirm the determinant role of the 4. Sanz-Leon P, Knock SA, Woodman Domide L, Mersmann J, McIntosh AR,
connectome in spatial seizure propagation, we performed several Jirsa VK: The Virtual Brain: a simulator of primate brai network dynamics.
surrogate analysis with other neural mass models (e.g. FitzHugh-Nagumo Frontiers in Neuroinformatics 2013, 1-23.
model), connectivity of control subjects and generic connectivities such
as shuffled connectivities, random and small-world networks, again
evaluated against clinical data. Real connectomes always performed
better than generic connectivities. The connectome particular structure of P111
a patient was often but not always better to predict seizure propagation Adaptive control of ventilation using electrical stimulation in a
than connectome of controls. biomechanical model
In conclusion, our results show that large-scale white matter tracts play an Brian K Hillen1*, James J Abbas2, Adeline Zbrzeski3, Sylvie Renaud3,
important role in the propagation of epileptic seizures. Better understanding Ranu Jung1
1
of their exact role can help to significantly improve the success rate of Department of Biomedical Engineering, Florida International University,
surgical resections for epileptic patients. Miami, FL, 33174, USA; 2School of Biological and Health Systems
Acknowledgements: This research has been supported by the James S. Engineering, Arizona State University, Tempe, AZ, 85287, USA; 3IMS
McDonnell Foundation. Laboratory, CNRS UMR 5218,Institut Polytechnique de Bordeaux, Talence,
References 33405, France
1. Spencer SS: Neural networks in human epilepsy: evidence of and E-mail: Brian.Hillen@fiu.edu
implications for treatment. Epilepsia 2002, 43:219-227. BMC Neuroscience 2015, 16(Suppl 1):P111
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Cervical spinal cord injury (SCI) causes loss or impairment of control of
respiratory muscles. Life-sustaining ventilation can be provided by
mechanical ventilators (which have numerous side effects) or open-loop
electrical stimulation respiratory pacing systems [1]. The use of adaptive
control strategies in respiratory pacing systems can simplify initial setup
procedures and allow the system to adjust stimulation values to account
for changes due to muscle fatigue and/or respiratory demand. We have
implemented a neural network based adaptive controller [2] with a
biomechanical model of human ventilator dynamics and diaphragm
stimulation in Simulink/SimMechanics/ Matlab. The adaptive controller
uses sensor information to automatically determine a stimulation pattern
that will produce a pre-specified desired lung volume trajectory. The
controller uses a two-stage pattern generator/pattern shaper (PG/PS)
structure which has successfully controlled leg movements in human
subjects [3] and rats [4] using neuromuscular electrical stimulation. The
biomechanical model incorporates a physiologically realistic Hill-type
muscle model and a damped spring with non-linear compliance using
published parameters for muscle geometry [5]. The parameters of the
biomechanical model (muscle mass and lung damping) were varied
+/-20% to simulate variation across a population to yield 9 sets of
parameters. The quality of control and rate of adaptation achieved by the
PG/PS controller were quantified by assessing the tracking error (difference
between the actual and desired volume patterns) and the number of
cycles needed to reach 5% error. Controller parameters were initialized to
provide a nominal degree of ventilation during initial breaths. For each set
of biomechanical parameters, the controller adapted stimulation values to
achieve the same desired volume trajectory without any modification of
initial controller values (Figure 1) and to achieve less than 5% error in 1-10
(mean 5.4 2.6) cycles. This adaptive strategy will be investigated further
in simulation as well as hardware implementations for testing in animal
models.
Acknowledgements: This work was supported by NIH R01-NS086088
and ANR-13-NEUC-0001-01.
Figure 1(abstract P111) A: Stimulation output for each biomechanical model tested for the first 10 breaths. B: Lung volume trajectories for each
biomechanical model. The trial on the system with the weakest muscle and greatest load (damping) is shown in red. Note that all models achieved the
desired lung volume but required different stimulation patterns
Page 81 of 200
References
1. Hillen B, Jung R: Peripheral Nerve Interface Applications, Respiratory
Pacing. Encyclopedia of Computational Neuroscience Berlin Heidelberg:
Springer-Verlag: Jaeger D, Jung R 2013.
2. Stites EC, Abbas JJ: Sensitivity and versatility of an adaptive system for
controlling cyclic movements using functional neuromuscular
stimulation. IEEE Trans Biomed Eng 2000, 47:1287-1292.
3. Riess J, Abbas JJ: Adaptive neural network control of cyclic movements using
functional neuromuscular stimulation. IEEE Trans Rehabil Eng 2000, 8:42-52.
4. Fairchild MD, Kim SJ, Iarkov A, Abbas JJ, Jung R: Repetetive hindlimb
movement using intermittent adaptive neuromuscular electrical
stimulation in an incomplete spinal cord injury rodent model. Exp Neurol
2010, 223:623-633.
5. Hillen B, Jung R: Computational model of human ventilation for electrical
stimulation following cervical spinal cord injury. BMC Neuroscience 2014,
15:P133.
P112
Predictable implications of random photon absorption for
photoreceptors gain control
Zhuoyi Song1,3*, Yu Zhou2, Mikko Juusola3,4
1
Centre for Mathematics, Physics and Engineering in the Life Sciences and
Experimental Biology (CoMPLEX), University College London, London, WC1E
6BT, UK; 2School of Computing, Engineering and Physical Sciences,
University of Central Lancashire, PR1 2HE, UK; 3Department of Biomedical
Science, University of Sheffield, S10 2TN, UK; 4State Key Laboratory of
Cognitive Neuroscience and Learning, Beijing Normal University, Beijing
100875, China
E-mail: zhuoyi.song@ucl.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P112
Light intensities change enormously in terrestrial environments, from
murky starlit night to scorching daylight [1]. Photoreceptors of land
animals have evolved with specialized photon absorption structures and
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
adaptive mechanisms to cope with this large input range. They can
effectively encode the vastly varying light intensities to macroscopic
voltage responses, capturing the temporal structure of natural contrast
changes within their limited output range. However, if photoreceptors
were purely photon counters, counting every single photon that hits
them, their outputs would readily saturate at bright daylight. To lessen
this problem, it has been suggested that sublinear summation in
quantum bump production (quantum-gain nonlinearity) may reduce the
bump/photon gain at the instances when multiple photons hit the same
photon-sampling-unit (multi-photon-hits) [2]. Here, we quantify the
contribution of this nonlinearity to light adaptation, using a Random
Photon Absorption Model for microvillar photoreceptors. We show that the
quantum-gain nonlinearity affects only marginally ( 1%) photoreceptors
with many microvilli, such as those of flies. This is because, with tens of
thousands of photon-sampling-units, the probability of multiple photons
hitting on any one of them simultaneously is very low. However, this
nonlinear mechanism is predicted to affect a photoreceptors encoding
more, if the cell has fewer microvilli, especially when it faces a bright
daylight environment.
References
1. van Hateren JH: Processing of natural time series of intensities by the
visual system of the blowfly. Vision research 1997, 37(23):3407-3416.
2. Pumir A, Graves J, Ranganathan R, Shraiman BI: Systems analysis of the
single photon response in invertebrate photoreceptors. Proceedings of
the National Academy of Sciences of the United States of America 2008,
105(30):10354-10359.
P113
An increase in the extracellular potassium concentration can cause
seizures
Tianlin Ying1, David B Grayden1,2, Anthony N Burkitt1,2, Tatiana Kameneva1*
1
NeuroEngineering Laboratory, Department of Electrical and Electronic
Engineering, University of Melbourne, Melbourne, Victoria 3010, Australia;
2
Bionics Institute, Melbourne, Victoria 3002, Australia
E-mail: tkam@unimelb.edu.au
BMC Neuroscience 2015, 16(Suppl 1):P113
Introduction: Epilepsy is a neurological disorder characterized by recurrent
seizures. It is important for the development of patient treatments to
understand the mechanisms underlying this complex neurological disease.
Experimental data shows that an increase in the extracellular potassium
concentration, [K+]0, can support the generation of seizures and growth in
seizures frequency and propagation velocity [1]. It is unclear if seizures are
caused by the increase in [K+]0 or seizures cause the [K+]0 increase.
Methods: Using a single column model neural mass model of pyramidal
cell population [2], we explore how the change in [K+]0 affects the model
output. The model represents the activity of approximately 105 neurons
and has four interacting neural populations: pyramidal neurons, excitatory,
slow- and fast-inhibitory neurons. The model output is interpreted as
representing the recorded EEG signal. A static nonlinear function (a
sigmoid) of the model converts a postsynaptic potential into an average
spiking rate for each neural population. We fit data from [3] to different
forms of the sigmoid corresponding to different concentration of [K+]0.
Simulations were run with different sigmoid parameters while other
parameters of the model are fixed to produce normal background activity
(non-seizure dynamics). To classify the model output as a particular neural
activity type, a power spectral density analysis is used and the types were
compared to those in [2].
Results: Results show that a [K+]0 increase from 5 to 13 mM causes the
neural dynamics to transit from Type 1 (normal background activity) to
Type 6 (slow quasi-sinusoidal activity) to Type 3 (sustained discharge of
spikes) to Type 1, and then to Type 5 (low-voltage rapid activity).
Conclusions: Our results confirm that a [K+]0 increase can generate seizures.
This may have implications on the development of the effective treatments
for epilepsy patients. It is left for future research to investigate whether
changes in [K+]0 can affect the frequency or propagation velocity of seizures.
These investigations will be carried out using multiple interconnected
columns.
Acknowledgements: This research was supported by the Australian
Research Council (DE120102210). The Bionics Institute acknowledges the
Page 82 of 200
support it receives from the Victorian Government through its
Operational Infrastructure Support Program.
References
1. Jensen MS, Yaari Y: Role of intrinsic burst firing, potassium accumulation,
and electrical coupling in the elevated potassium model of
hippocampal epilepsy. J Neurophys 1997, 77:1224-1233.
2. Wendling F, Bartolomei F, Bellanger JJ, Chauvel P: Epileptic fast activity
can be explained by a model of impaired GABAergic dendritic
inhibition. Europ J Neurosci 2002, 9:1499-1508.
3. Zandt BJ, Visser S, van Putte MJAN, ten Haken B: A neural mass model
based on single cell dynamics to model pathophysiology. J Comput
Neurosci 2014, 37:549-568.
P114
Discrete cortical representations and their stability in the presence of
synaptic turnover
Bastian Eppler1,2*, Dominik Aschauer3, Simon Rumpel3, Matthias Kaschube1,2
1
Frankfurt Institute for Advanced Studies, 60438 Frankfurt, Germany;
2
Goethe-Universitt, 60438 Frankfurt, Germany; 3Johannes-Gutenberg-
Universitt, 55122 Mainz, Germany
E-mail: eppler@fias.uni-frankfurt.de
BMC Neuroscience 2015, 16(Suppl 1):P114
Population imaging in mouse auditory cortex revealed clustering of neural
responses to brief complex sounds: the activity of a local population
typically falls close to one out of a small number of observed states [1].
These clusters appear to group sets of auditory stimuli into a discrete set
of activity patterns and could thereby form the basis for representations of
sound categories. However, to be useful for the brain, such representations
should be robust against fluctuations in the underlying circuitry, which are
significant even in the absences of any explicit learning paradigm [2]. Here
we introduce a novel firing rate based circuit model of mouse auditory
cortex to study the emergence of the observed activity cluster states and
their structural stability in the presence of synaptic noise. We find that
generic random networks by virtue of their inhibitory recurrent
connectivity can group complex sounds spontaneously into essentially
discrete sets of activity states. Moreover, these states can display high
degrees of stability, even when modifying a substantial fraction of synaptic
connections, as long as the basic statistics of connectivity is maintained.
We use the insights gained from the analysis of our model to interpret
data gathered in a parallel effort, employing chronic two-photon imaging
of population activity in the auditory cortex of awake mice.
References
1. Bathellier B, Ushakova L, Rumpel S: Discrete neocortical dynamics predict
behavioral categorization of sounds. Neuron 2012, 76(2):435-49.
2. Loewenstein Y, Kuras A, Rumpel S: Multiplicative dynamics underlie the
emergence of the log-normal distribution of spine sizes in the
neocortex in vivo. J Neurosci 2011, 31(26):9481-8.
P115
Influence of recurrent interactions on texture processing in networks
with different visual map organizations
Hanna Kamyshanska*, Dmitry Bibichkov, Matthias Kaschube
Frankfurt Institute for Advanced Studies and Faculty of Computer Science
and Mathematics, Johann Wolfgang Goethe University, Frankfurt am Main,
60438, Germany
E-mail: kamyshanska@fias.uni-frankfurt.de
BMC Neuroscience 2015, 16(Suppl 1):P115
The functional architecture of the visual cortex displays marked differences
across mammalian species: in stark contrast to primates, in which the
preferred stimulus orientation forms an almost smooth map across the
cortical surface, in rodents a salt-and-pepper organization has been
observed [1]. It is conceivable that the organization of preferred orientation
has an impact on the processing of visual input. Recently [2], we found that
in a biologically inspired object recognition system with a pure feed forward
network architecture, smooth orientation maps outperform the salt-and-
pepper organization in a texture recognition task. Here, we extend this work
to study the effect of recurrent connections on neuronal response
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
properties and texture recognition, comparing the two types of cortical
architectures.
Our model is a recurrent single-layer rate network. The feed forward
connections to each neuron are properly oriented Gabor-filters. Recurrent
connections between two neurons depend both on the spatial distance
between these neurons and on their orientation selectivities. Inspired by [3],
the angle-dependent interaction function is weighted by the product of
selectivities of both neurons and enables excitation between cells with
similar preferred orientations and inhibition between cells with orthogonal
ones. We also study the effects of purely distant-dependent recurrent
connectivity of Mexican-hat type, with spatial extent related to local column
spacing in case of smooth map layout. We design a network for the salt-
and-pepper organization in an analogous way, assuming the same spatial
extent of connectivity and its tuning-dependence as observed in mouse
visual cortex [4].
Varying the strength and selectivity of recurrent versus feed-forward
connections, we first explore the influence of recurrence on the orientation
selectivity. For that purpose, we drive the network with oriented gratings
to reconstruct the selectivity from the activities. In agreement with [5] we
observe sharpening of the orientation tuning of the network by recurrent
interactions, such that oriented stimuli can be well discriminated even with
weak feed-forward tuning. We further study the role of recurrent
connections in processing more complex stimuli. We present visual
textures to the model, then feed the responses into a classifier (linear SVM)
that learns to predict a class label. This allows us to study how differences
in feed-forward and recurrent connections impact texture classification,
and to compare the orientation-preference map and salt-and-pepper
organization in texture recognition tasks.
References
1. Kaschube M: Neural maps versus salt-and-pepper organization in visual
cortex. Curr Opin Neurobiol 2014, 24:95-102.
2. Bauer F, Kaschube M: Processing textures in a smooth visual map and a
salt-and-pepper organization. Bernstein Conference 2013.
3. Blumenfeld B, Bibitchkov D, & Tsodyks M: Neural network model of the
primary visual cortex: From functional architecture to lateral connectivity
and back. J Comput Neurosci 2006, 20(2):, 219-241.
4. Ko H, Hofer SB, Pichler B, Buchanan KA, Sjstrm PJ, Mrsic-Flogel TD:
Functional specificity of local synaptic connections in neocortical
networks. Nature 2011, 473(7345):, 87-91.
5. Ben-Yishai R, Bar-Or RL, Sompolinsky H: Theory of orientation tuning in
visual cortex. Proc Natl Acad Sci U S A 1995, 92:, 3844-3848.
P116
A Newton-based shooting method to find synaptic threshold in active cables
William L Kath
Departments of Applied Mathematics and Neurobiology, Northwestern
University, Evanston, IL 60201, USA
E-mail: kath@northwestern.edu
BMC Neuroscience 2015, 16(Suppl 1):P116
The integration of synaptic inputs in a neuron can be nonlinear not just at
the axon, but also locally in the dendrites if they are imbued with active
Figure 1(abstract P116) Threshold Fitzhugh-Nagumo voltage profile in a branched cable for synapse position just past the branch point (left;
blue is main and one daughter branch; red is the other) and critical conductance value as a function of synapse position in main and one
daughter branch (right). The branch point is at position = 40
Page 83 of 200
voltage-gated ion channels. For example, CA1 pyramidal neurons have
high densities of sodium and potassium and currents in their dendrites,
and these densities can vary substantially in the arbor [1,2]. Such
nonlinearities can lead to compartmentalized responses to inputs [3], with
branches acting as individual nonlinear units in which dendritic spikes
occur. A cell can thus function as a multi-layered network with the soma as
final output. This motivates determining when a given set of synaptic
inputs is large enough to generate a local dendritic spike, or, alternatively,
determining the synaptic conductance value(s) at threshold for producing
a spike.
Above- and below-threshold conditions are known to be separated by
the threshold or critical surface [4]. Here it is shown that the synaptic
conductance leading to a threshold solution can be found by modifying
Newton methods developed to find steady-state solutions in fluid
mechanics [5]. Consider a general form of the cable equation,
u
= L(x)u + N (u, x) + gs Gs (t, x)(urev u)
t
Here u represents the voltage and any gating variables. The first term on
the right models the diffusive part of the cable equation, and the next
term the nonlinearities from any active voltage-gated ion channels. The
last term represents synaptic conductances at points in the dendritic tree
with overall strength gs. An unstable threshold solution and accompanying
synaptic strength can be found using 1) a preconditioned version of the
steady-state cable equation combined with constraints requiring the
difference between a shooting solution from rest, u(T,x;gs), and the critical
surface to be perpendicular to the single unstable eigenvector associated
with the critical surface. The overall procedure finds the value of gs leading
to a solution that asymptotes to the critical surface as t goes to .
Example results from the method are shown in Figure 1. These are for the
Fitzhugh-Nagumo model without recovery in a y-branched cable
morphology where the daughter branches have half the diameter of the
main dendrite. One observes the expected drop of threshold conductance
as the synapse moves from the main to daughter branch, but this
transition is not monotonic as for a passive cable because of the nonlinear
active conductance in the model. Discovering phenomena such as this
would be quite laborious without employing a Newton-based method.
Acknowledgements: WLK was supported by NIH grant 1R01NS077601 &
the Janelia Research Campus Visiting Scientist Program.
References
1. Spruston N: Pyramidal neurons: dendritic structure and synaptic
integration. Nat. Rev. Neurosci 2008, 9(3):206-21.
2. Johnston D, Magee JC, Colbert CM, Cristie BR: Active properties of
neuronal dendrites. Annual Rev. Neurosci 1996, 19:165-186.
3. Polsky A, Mel BW, Schiller J: Computational subunits in thin dendrites of
pyramidal cells. Nat. Neurosci 2004, 7(6):621-7.
4. Mckean HP, Moll V: Stabilization to the standing wave in a simple
caricature of the nerve equation. Comm. Pure Appl. Math 1986,
39(4):485-529.
5. Tuckerman LS, Huepe C, Brachet ME: Numerical methods for bifurcation
problems. Instabilities and non-equilibrium structures IX O. Descalzi, J.
Martinez, and S. Rica 2004, 9.
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 84 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

such models to match experimental data remains challenging. Here we

P117
Fully-automated multi-objective optimization for fitting a neuronal
model with real morphology
Aushra Abouzeid1*, Nelson Spruston2, William Kath1
1
Engineering Sciences and Applied Mathematics, Northwestern University,
Evanston, IL 60208, USA; 2Howard Hughes Medical Institute, Janelia Research
Campus, Ashburn, VA 20147, USA
E-mail: aushra.abouzeid@northwestern.edu
BMC Neuroscience 2015, 16(Suppl 1):P117
Morphologically realistic models have successfully been used to elucidate
many complex mechanisms in neuronal dendrites. However, the tuning of
introduce a fully automated parameter optimization methodology that
uses the Python programming language to control the NEURON simulator
in parallel on a high performance computing cluster.
Using targeted experimental protocols, including sub- and supra-threshold
somatic as well as dendritic voltage recordings, we constrain a model
hippocampal CA1 pyramidal cell built with a complete reconstructed
morphology. The optimization is performed using the non-dominated
sorting genetic algorithm (NSGA-II), and model fitness is evaluated by
directly comparing the simulated and recorded voltage traces. In order to
impose minimal a priori assumptions, we use a multi-objective framework,
which tunes all of the free parameters with respect to all of the
experimental objectives simultaneously. Furthermore, the multi-objective

Figure 1(abstract P117) Clickable interface for browsing the Pareto front of solutions produced by our multiobjective optimization platform.
Top row: Scatterplot of error scores across a 6-dimensional error space. The abscissa and ordinal represent errors in the objectives shown below in the
same column. A highlighted point appears in each scatterplot (white circle with red center) to indicate errors produced by a single model instance across
each of the six objectives. Bottom two rows: Experimental traces (red) and model output (blue). The right-most column depicts dendritic voltage
attenuation as a function of distance from soma as measured in dual patch-clamp recordings along the main apical dendrite
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
approach avoids the pitfalls of overfitting, because the algorithm produces a
diverse family of solutions on the so-called Pareto-optimal front. To facilitate
model selection, we have developed a clickable interface for visually
browsing the set of optimal solutions, which permits the explicit and rapid
identification of trade-offs among the fitting objectives and the biophysical
parameters that govern variability in the solution set.
P118
Granule cell excitability mediates gamma and beta oscillations in a
model of the dendrodendritic microcircuit
Boleslaw L Osinski1,2*, Leslie M Kay2,3
1
Biophysical Sciences, University of Chicago, Chicago, IL 60637, USA;
2
Institute for Mind and Biology, University of Chicago, Chicago, IL 60637,
USA; 3Department of Psychology, University of Chicago, Chicago, IL 60637,
USA
E-mail: boleszek@uchicago.com
BMC Neuroscience 2015, 16(Suppl 1):P118
Odors evoke gamma (60 - 100 Hz) and beta (20 - 30 Hz) oscillations in the
local field potential (LFP) of the rat olfactory bulb (OB). These oscillations
arise from activity in the dendrodendritic microcircuit between excitatory
mitral cells (MCs) and inhibitory granule cells (GCs) [1]. When cortical
feedback inputs to the OB are blocked, beta oscillations are extinguished
while gamma oscillations persist [2]. Much of this cortical feedback targets
inhibitory interneurons in the GC layer and regulates the excitability of GCs
[3], which suggests a causal link between the emergence of beta
oscillations and the GC excitability. We investigate the effect that GC
excitability has on network oscillations in a biophysical model of the MC-
GC dendrodendritic network with graded inhibition. When GC excitability
is low, there is transient activation of NMDAR channels by AMPARs, which
produces fast inhibitory pulses in the gamma frequency range. When GC
excitability is increased, the activation of NMDARs and other VDCCs is
prolonged, allowing the slow decay time constants of these channels to
drive beta frequency oscillations. The power of the beta oscillation peaks
when inhibitory and excitatory currents onto MCs are balanced, which
could explain the relationship between beta power and odor volatility
measured experimentally [4].
References
1. Lagier S, Carleton A, Lledo P-M: Interplay between local GABAergic
interneurons and relay neurons generates gamma oscillations in the rat
olfactory bulb. J. Neuroscience 2004, 24:4382-92.
2. Pressler RT, Inoue T, Strowbridge BW: Muscarinic receptor activation
modulates granule cell excitability and potentiates inhibition onto mitral
cells in the rat olfactory bulb. J. Neuroscience 2007, 27:10969-81.
3. Neville KR, Haberly LB: Beta and gamma oscillations in the olfactory
system of the urethane-anesthetized rat. J. Neurophysiology 2003,
90:3921-30.
4. Lowry CA, Kay LM: Chemical factors determine olfactory system beta
oscillations in waking rats. Journal of neurophysiology 2007, 98:394-404.
P119
Distinct and competing interneuron populations can generate fast and
slow gamma in oscillatory models of CA1
Stephen L Keeley1*, Andre A Fenton1, John Rinzel1,2
1
Center for Neural Science, New York University, New York, NY 10003, USA;
2
Courant Institute of Mathematical Sciences, New York University, New York,
NY 10012, USA
E-mail: StephenLKeeley@gmail.com
BMC Neuroscience 2015, 16(Suppl 1):P119
Gamma oscillations are widely observed in the mammalian brain and are
important markers for cognition and attention [1,2]. In CA1 of the
hippocampus of freely moving rats, power in one of two distinct oscillatory
bands in the gamma regime (fast gamma and slow gamma) is predominantly
present at a given moment of time [3]. Here, we demonstrate that models of
networks with competing interneuron populations with different post-
synaptic effects can create distinct oscillatory regimes that mimic the
observed oscillations of CA1. Our network formulation reflects the following
facts: 1) The duration of post-synaptic effect of an interneuron strongly
influences the frequency in biophysical models of gamma oscillations [4].
Page 85 of 200
2) The primary CA1 inputs from CA3 and the entorhinal cortex (EC)
preferentially innervate interneurons of different subtype with different post-
synaptic durations [5,6].
We show that a firing rate model with competing interneuron populations
with different post-synaptic time-constants is sufficient to generate slow
and fast gamma oscillations. We conclude that mutual inhibition between
the modeled interneuron populations permits switching in a bistable
regime between distinct fast and slow gamma states. We also find similar
behavior in spike-based network models. Our models explicitly predict the
following about CA1: 1) Different interneurons innervated by different
upstream regions phase-lock to different gamma states. 2) One population
of interneurons is silenced, and another is active during fast and slow
gamma events. 3) Mutual inhibition between interneuron populations is
necessary for spontaneous switching of gamma state. Using experimental
electrophysiological data from awake behaving rodents, we find
interneurons that satisfy conditions 1 and 2, and we show putative fast
and slow gamma interneurons categorized by their tendency to fire and
phase-lock with oscillatory events as measured by a nearby local field
potential.
Our 3-population firing rate model is schematized in Figure 1A. The
dynamic variables are synaptic currents of an excitatory, fast inhibitory (IF)
and slow inhibitory (Is ) population; the firing rates are instantaneous
functions of total input current. Fast excitation that interacts with
inhibitory subpopulations supports oscillations. This interaction engages
either one or both inhibitory subpopulations depending on I S - I F
connectivity and input balance (Example in Figure 1B). This network
oscillates at biophysically realistic frequencies given biophysically realistic
network parameters. The fast inhibitory population, IF and slow inhibitory
population, I S have post-synaptic time-constants of 5ms and 15ms,
respectively. These roughly capture the diversity of post-synaptic
inhibitory current time-courses of interneurons of different subtypes
measured in CA1 [6]. Our firing rate model demonstrates that with
sufficient mutual inhibition between inhibitory populations, the oscillating
network bifurcates into two stable regimes that oscillate at roughly the
same frequencies as the observed fast and slow gamma oscillations [3,7].
Previous experimental work suggests these two gamma oscillations reflect
different information processing modes in the learning and memory
system [7]. Our models provide a mechanistic understanding of these
modes and posit a new oscillatory role for distinct interneurons in CA1.
Moreover, our models describe general oscillatory behavior in networks
with distinct interneuron populations.
References
1. Fries P, Reynolds JH, Rorie AE, Desimone R: Modulation of oscillatory
neuronal synchronization by selective visual attention. Science 2001,
291(5508):1560-3.
2. Baar E, Baar-eroglu C, Karaka S, Schrmann M: Gamma, alpha, delta, and
theta oscillations govern cognitive processes. Int J Psychophysiol 2001,
39(2-3):241-8.
3. Belluscio MA, Mizuseki K, Schmidt R, Kempter R, Buzski G: Cross-frequency
phase-phase coupling between and g oscillations in the hippocampus.
J Neurosci 2012, 32(2):423-35.
4. Jefferys JG, Traub RD, Whittington MA: Neuronal networks for induced 40
Hz rhythms. Trends Neurosci 1996, 19(5):202-8.
5. Gulys AI, Megas M, Emri Z, Freund TF: Total number and ratio of
excitatory and inhibitory synapses converging onto single interneurons
of different types in the CA1 area of the rat hippocampus. J Neurosci
1999, 19(22):10082-97.
6. Maccaferri G, Roberts JD, Szucs P, Cottingham CA, Somogyi P: Cell surface
domain specific postsynaptic currents evoked by identified GABAergic
neurones in rat hippocampus in vitro. J Physiol (Lond) 2000,
524(Pt 1):91-116.
7. Colgin LL, Denninger T, Fyhn M, Hafting T, Bonnevie T, Jensen O,
Moser MB, Moser EI: Frequency of gamma oscillations routes flow of
information in the hippocampus. Nature 2009, 462(7271):353-7.
P120
Modeling spontaneous activity across an excitable epithelium: support
for a coordination scenario of early neural evolution
Oltman O de Wiljes1*, Ronald AJ van Elburg2, Michael Biehl3, Fred A Keijzer1
1
Department of Theoretical Philosophy, Groningen University, Groningen, the
Netherlands; 2Institute of Artificial Intelligence, Groningen University,
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 86 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

Figure 1(abstract P119) A. Connectivity scheme for firing-rate two-gamma model. CA3 and EC denote inputs, E denotes excitatory population, and
IF and IS denote the interneuron populations with brief and long post-synaptic effects, respectively. B. Graph shows the oscillation frequency of the
noise-free network across a range of input balance to IS and IF. Blue region indicates bistable regime with co-existing fast and slow oscillatory states. Such
a bistable regime only exists with high IS - IF connectivity. C. Time courses of each populations synaptic variable show spontaneous switching between
fast and slow gamma states due to additive noise in inputs to IS and IF

Groningen, the Netherlands; 3Johann Bernoulli Institute for Mathematics and Our results show that basic internal coordination as proposed by the skin
Computer Science, Groningen University, Groningen, the Netherlands brain thesis could have arisen in this potential nervous system precursor,
E-mail: o.o.de.wiljes@rug.nl providing support that this configuration may have played a role as a
BMC Neuroscience 2015, 16(Suppl 1):P120 proto-neural system and requires further investigation.

The reason why nervous systems first arose is an open question. Internal
coordination models hold that nervous systems evolved initially as a
device to coordinate internal activity, enabling multicellular effectors.
They stress the use of multicellular contractility as an effector for motility:
some sort of coordinative structure would have been necessary to have
multicellular effectors in the first place. A recent example of such a
model, the skin brain thesis, suggests that excitable epithelia using
chemical signaling are a potential candidate as a nervous system
precursor.
We developed a computational model and a measure for whole body
coordination to investigate the coordinative properties of such excitable
epithelia. Using this measure we show that excitable epithelia can
spontaneously exhibit body-scale patterns of activation (see Figure 1.).
Relevant factors determining the extent of patterning are the noise level
for exocytosis, relative body dimensions, and body size. In smaller bodies
whole-body coordination emerges from cellular excitability and
bidirectional excitatory transmission alone.
P121
Mechanisms of spikelet generation in cortical pyramidal neurons
Martina Michalikova1*, Michiel Remme1, Richard Kempter1,2
1
Institute for Theoretical Biology, Department of Biology, Humboldt-
Universitt zu Berlin, 10115 Berlin, Germany; 2Bernstein Center for
Computational Neuroscience, 10115 Berlin, Germany
E-mail: martina.michalikova@hu-berlin.de
BMC Neuroscience 2015, 16(Suppl 1):P121
Spikelets are brief, spike-like depolarizations of small amplitude (< 20 mV)
that can be measured in somatic intracellular recordings. Prominent
spikelet activity was demonstrated in hippocampal CA1 pyramidal
neurons in awake behaving [1,2] and anesthetized animals [3]. However,
spikelets are rarely observed in vitro, and basic mechanisms underlying
their generation in pyramidal neurons are not well understood.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P120) Wave patterns on a long tube (length: 32
cells, circumference: 8 cells, noise rate: 0.1 Hz). (A) Network
geometry: the scissors indicate the line at which the tube is cut for
presentation in B. (B) Snapshots of network activity during 4 ms
intervals in an illustrative phase of the dynamics. The wave fronts
propagating transversely collide with each other, causing extinction due
to the refractory period. Subsequently, the remaining wave-fronts
propagating longitudinally dominate the dynamics and the North-South
wave-front orientation dominates. (C) Temporal development of
different wave-front orientations. Snapshot time markings are consistent
with those in B
Page 87 of 200
Here we investigate the emergence of spikelets using mathematical
analysis and numerical simulations of compartmental single-neuron
models. Somatic spikelets are produced in the models upon orthodromic
(somatic) stimulation. We find that spikelet occurrence depends on three
main factors: A) Activation voltages of somatic channels need to be
larger by several millivolts (~10 mV) than activation voltages of axonal
sodium channels. B) The spike initiation zone (axon initial segment, AIS)
has to be electrically segregated from the soma. C) The impedance
mismatch between soma and AIS needs to be sufficiently large. In this
way, weak orthodromic stimuli can trigger APs at the AIS that fail to
activate somatic sodium channels and manifest as somatic spikelets.
Stronger stimuli lead to full-size APs at the soma, either through axonal
APs that backpropagate to the soma (shouldered APs) or direct somatic
AP generation (full-blown APs).
Through analysis and simulations we isolated the cell parameters that
allow for spikelet generation and identified possible causes of spikelet
absence in in vitro preparations: First, the dendritic current sink in vitro is
diminished due to dendritic pruning in slices. Next, the fraction of
sodium channels usually available for (somatic) spiking is larger in vitro due
to the overall lower spiking activity and lower resting membrane potential.
Finally, the difference in activation voltages between somatic and axonal
sodium channels under in vitro conditions might be smaller than under
in vivo conditions as the activation voltage of sodium channels might be
controlled by neuronal activity, which is typically much higher in vivo than
in vitro.
Conclusions: In our models, somatic spikelets represent APs that are only
propagated down the axon, but are not backpropagated to the soma and
the dendrites. Consequently, such a mechanism might be involved in the
control of dendritic plasticity and/or in the homeostatic regulation of
somato-dendritic firing rates without affecting the axonal output of a neuron.
Acknowledgements: This work was supported by the Einstein Foundation
Berlin and the German Federal Ministry of Education and Research
(01GQ0901, 01GQ1001A, 01GQ0972).
References
1. Epsztein J, Lee AK, Chorev E, Brecht M: Impact of spikelets on
hippocampal CA1 pyramidal cell activity during spatial exploration.
Science 2010, 327(5964):474-477.
2. Harvey CD, Collman F, Dombeck DA, Tank DW: Intracellular dynamics of
hippocampal place cells during virtual navigation. Nature 2009,
461(7266):941-946.
3. Chorev E, Brecht M: In vivo dual intra- and extracellular recordings
suggest bidirectional coupling between CA1 pyramidal neurons.
J Neurosci 2013, 33(11):4815-4824.
P122
Spiking network modeling of neuronal dynamics in individual rats
John S Choi1, Rosemary J Menzies2, Salvador Dura-Bernal1, Joseph T Francis1,
William W Lytton1, Cliff C Kerr1,2,3*
1
Department of Physiology and Pharmacology, SUNY Downstate Medical
Center, Brooklyn, NY 10029, USA; 2Complex Systems Group, School of
Physics, University of Sydney, Sydney, NSW 2006, Australia; 3Centre of
Excellence for Integrative Brain Function, University of Sydney, Sydney, NSW
2006, Australia
E-mail: cliff@thekerrlab.com
BMC Neuroscience 2015, 16(Suppl 1):P122
Many researchers base their neuronal models on experimental data, but
few systematically calibrate to it, and even fewer use unpooled data from
multiple subjects. Some (partial) exceptions are [1], where one model
parameter was calibrated to data pooled across four macaques; and [2],
where five model parameters were calibrated to unpooled data from 292
humans. To our knowledge, no one has previously calibrated spiking
network models to individual subjects - likely because a single model
iteration typically takes considerable computational time, compounded by
the formidable number of iterations required to perform optimizations on
a statistically meaningful number of subjects.
In this study, we calibrate and validate a mesoscopic spiking network model
against data from individual rats, then use these fits to infer differences in
the rats physiologies. We recorded data from microelectrode arrays
implanted in the somatosensory cortices of nine male rats, each of whom
received touch stimuli to his left forepaw. The spiking network models
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
consisted of 20,000 Izhikevich neurons, representing a 2x2 mm patch of
cortex sampled at approximately 10% true cell density, with cell types
(15 across six layers) and connectivities based on empirical data. We
calibrated key model parameters (including connection probability and
weight, tonic background activity, and the ratio of thalamic to cortical input)
to experimental data (including average firing rates and coefficients of
variation) using a nonlinear optimization algorithm [3]. The calibrations were
validated using the exponents of the local field potential (LFP) power
spectra from 5-50 Hz, which were not used for calibration. Changes in the
information processing properties of the (simulated) networks were then
quantified via interlaminar Granger causality.
Experimentally, we found large differences between subjects. Cortical firing
rates varied from 3.8 Hz to 19.6 Hz (median 11.9 Hz); coefficients of
variation, 0.4 to 1.2 (median 0.9); and peristimulus time histogram peak
amplitudes, 48 Hz to 73 Hz (median 62 Hz). Inter-subject differences were
significantly greater than intra-subject differences across both sessions and
electrodes. The calibrated models reproduced experimental data with an
average mismatch of 16%, compared to 30% when uncalibrated.
Experimentally, the LFP power spectra exponents varied from -1.4 to
-2.7 (median -2.2). While the models had uniformly steeper exponents
(range -2.8 to -3.6), calibration reduced mismatch in eight of the nine
subjects. Inter-subject differences could be largely accounted for by
differences in average synaptic connection probability (which varied by a
mean of 24% between subjects, compared to <8% for other fitted
parameters). Parameter differences produced changes in how these
networks process information - for example, there were significant and
complex differences in the patterns of interlaminar information flow
between subjects, including reversals in the dominant direction of
information flow for some layer pairs.
In summary, we found that (1) significant differences exist between
individual subjects, (2) these differences can be captured by calibrating
spiking network models to data from each individual, and (3) modeling
suggests that these differences affect how individuals process information
across cortical layers.
References
1. Rasch MJ, Schuch K, Logothetis NK, Maass W: Statistical comparison of
spike responses to natural stimuli in monkey area V1 with simulated
responses of a detailed laminar network model for a patch of V1.
J Neurophysiol 2011, 105(2):757-78.
2. Kerr CC, Kemp AH, Rennie CJ, Robinson PA: Thalamocortical changes in
major depression probed by deconvolution and physiology-based
modeling. Neuroimage 2011, 54(4):2672-82.
3. Kerr CC, Smolinski TG, Dura-Bernal S, Wilson DP: Optimization by Bayesian
adaptive locally linear stochastic descent , Under review; available at http://
thekerrlab.com/pubs/ballsd.pdf..
P123
T-type calcium channels trigger a hyperpolarization induced
afterdepolarization in substantia nigra dopamine neurons
Rebekah C Evans*, Zayd M Khaliq
NINDS, NIH, Bethesda, MD 20892, USA
E-mail: rebekah.evans@nih.gov
BMC Neuroscience 2015, 16(Suppl 1):P123
Dopamine neuron dendrites integrate synaptic information arriving from a
diverse set of cell classes including excitatory and inhibitory neurons. The
majority of synapses formed onto substantia nigra (SNc) dopamine neurons
(>70 %) arrive from inhibitory GABAergic neurons [1]. Activation of GABA
receptors pauses tonic firing and hyperpolarizes cells. Although the pause in
dopamine neuron firing has behavioral relevance during reward omission
[2], the hyperpolarization may also enable burst firing through disinhibition
[3] or engagement of voltage-gated ion channels [4]. However, it is not clear
how hyperpolarization influences dendritic integration. Here we use detailed
single-cell computational modeling, patch-clamp electrophysiology, and
two-photon calcium imaging to investigate the dendritic response to
hyperpolarization. We find that at hyperpolarized potentials, some
dopamine neurons respond to brief somatic current injections with a long-
lasting depolarizing plateau that is accompanied by a large calcium
transient in the dendrites. Because this depolarizing plateau requires prior
hyperpolarization, we will refer to it as a hyperpolarization induced
afterdepolarization (HI-ADP).
Page 88 of 200
Using electrophysiology and calcium imaging, we found that
pharmacological block of T-type calcium currents with TTA-P2 completely
eliminated the dendritic calcium transient and greatly reduced the size of
the HI-ADP. In contrast, blocking L-type calcium channels with nifedipine
did not significantly alter the dendritic calcium or size of the HI-ADP. To
further investigate the involvement of dendritic T-currents in generating
the HI-ADP, we developed a multi-compartmental, multi-channel model of
an SNc dopamine neuron in Genesis simulation software. This model
contains a spherical soma with two dendritic trees extending from it, each
containing identical primary, secondary and tertiary branches. Our
simulations showed that a high density of dendritic T-type calcium
channels is critical to the generation of the HI-ADP. Specifically, we found
that reducing the T-type channel conductance by half throughout the
entire cell completely abolished the HI-ADP. On the other hand, removing
T-type channels from only one dendritic tree slightly reduced, but did not
eliminate the HI-ADP. These simulations show that a high, localized density
of T-type channels is more important to the generation of the HI-ADP than
the total number of T-type channels in the cell.
Further simulations predict that the tightly coupled, electrotonically
compact dendrites characteristic of SNc dopamine neurons are also
necessary for the production of the HI-ADP. In particular, reducing the
membrane input resistance of the model cell eliminated the HI-ADP.
Confirming this prediction, our experimental data show that reducing the
input resistance of SNc dopamine neurons through activation of G-protein
coupled inwardly-rectifying potassium channels (GIRKs) abolished the
HI-ADP and reduced the concomitant dendritic calcium transient.
In conclusion, we have shown that T-type calcium channels and
electrotonically compact dendrites are essential for generating a HI-ADP in
SNc dopamine neurons. These HI-ADPs represent an interesting response
to hyperpolarization that may be unique to SNc dopamine neurons having
the specific combination of high T-type channel density and tight
electrotonically compact dendrites. This particular combination of
characteristics may allow SNc dopamine neurons to respond to inhibition
or the release of inhibition with an increased ability to burst.
References
1. Henny P, Brown MTC, Northrop A, Faunes M, Ungless MA, Magill PJ,
Bolam JP: Structural correlates of heterogeneous in vivo activity of
midbrain dopaminergic neurons. Nat Neurosci 2012, 15:613-619.
2. Schultz W, Dayan P, Montague PR: A neural substrate of prediction and
reward. Science 1997, 275:1593-1599.
3. Lobb CJ, Wilson CJ, Paladini CA: A dynamic role for GABA receptors on
the firing pattern of midbrain dopaminergic neurons. J Neurophysiol
2010, 104:403-413.
4. Tateno T, Robinson HPC: The mechanism of ethanol action on midbrain
dopaminergic neuron firing: a dynamic-clamp study of the role of I(h)
and GABAergic synaptic integration. J Neurophysiol 2011, 106:1901-1922.
P124
Assessing performance of directed functional connectivity measures in
the presence of common source
Jisung Wang, Heonsoo Lee*, Seunghwan Kim
Physics Department, Pohang University of Science and Technology, Pohang,
South Korea
E-mail: beafool@postech.ac.kr
BMC Neuroscience 2015, 16(Suppl 1):P124
The mixing problem of electroencephalography (EEG) in the presence of
common source could affect directed functional connectivity measures,
resulting in an incorrect directionality of information flow between two
signals. Here we introduce directed Weighted Phase Lag Index (dWPLI), a
signed version of Weighted Phase Lag Index (WPLI) and compare this
measure to Granger Causality (GC), Symbolic Transfer Entropy (STE), Phase
Slope Index (PSI), and directed Phase Lag Index (dPLI) under common
source effects. Robustness of the measures was tested in both analytic and
simulating ways.
For the simulated time series, signals were generated from unidirectionally
coupled autoregressive model and linearly mixed to achieve volume
conduction and common noise effects.
As expected from the analytic calculation, dPLI and dWPLI were unaffected
by the volume conduction while PSI, GC and STE were largely affected.
(Figure 1A). The mean percentages of false identification were 14.007.66,
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 89 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

Figure 1(abstract P124) A. Directionalities of five directed connectivity measures under the volume conduction effect. Positive Directionality
Index (DI) implies correct directionality of information flow. As M (mixing ratio) increases, the DIs of GC, STE and PSI decrease. However, as expected in
analytic calculation, the dPLI and dWPLI are not affected. (DI is defined as a normalized information flow.) B. Directionalities of five directed connectivity
measures under the common noise effect. As M (noise-to-signal ratio) increases, the DIs signs of GC, STE and PSI change while those of dPLI and dWPLI
do not

22.354.78, 27.387.04, 1.991.36, and 1.631.28, for the GC, STE, PSI, dPLI,
and dWPLI, respectively.
For the common noise case, only dPLI and dWPLI had their signs
preserved. (Figure 1B). The mean percentages of false identification were
30.14 16.86, 49.74 16.63, 34.72 10.23, 2.68 1.90, and 2.34 1.55.
Furthermore, the dWPLI outperformed dPLI for the common noise case
(p <0.01, Wilcoxon signed-rank test) which was also predicted from
analytic calculation.
Present study shows the common source effects might lead to biased
results with incorrect directionality of information flow. Among the five
directed functional connectivity measures, dWPLI is much less affected by
common source effects.
P125
How slow K+ currents impact on spike generation mechanism?
Ryota Kobayashi1,2*, Katsunori Kitano3
1 590:4839-4858.
Principles of Informatics Research Division, National Institute of Informatics,
2-1-2 Hitotsubashi, Chiyoda-ku, Tokyo, Japan; 2Department of Informatics,
SOKENDAI (The Graduate University for Advanced Studies), 2-1-2
Hitotsubashi, Chiyoda-ku, Tokyo, Japan; 3Department of Human and
Computer Intelligence, Ritsumeikan University, 1-1-1 Nojihigashi, Kusatsu,
Shiga 525-8577, Japan
E-mail: r-koba@nii.ac.jp
BMC Neuroscience 2015, 16(Suppl 1):P125
Neuronal adaptation is the change in the responsiveness of a neuron over
time, and may improve coding information from an environment.
Adaptation originates from various factors, including single neurons,
synapses, and network dynamics. Here we investigate adaptation in a
responsiveness of a neuron. When a neuron received prolonged stimulation,
it initially responds with a high firing rate, and the firing rate decrease. This
is called spike-frequency adaptation, which is observed in most pyramidal
neurons in various animals. Spike-frequency adaptation is usually accounted
for by slow K + currents, for example, the M-type K + current (I M ) and
the Ca 2+ -activated K + current (I AHP ), and the conductance-based
(HodgkinHuxley type) models including the slow K + currents have
succeeded to reproduce the electro-physiological properties of a neuron [1].
The detailed biophysical mechanism underlying spike-frequency
adaptation may impact on the coding property of a neuron [2,3]. For
example, it was suggested that I M facilitates the spike-timing coding,
whereas I AHP improves the spike rate-coding [2] and I M increases the
response to low-frequency input signals, whereas I AHP decreases the
response to low-frequency signals [3].
Due to the complexity of the conductance-based models, it is not clear
how the slow K+ currents impact on spike generation mechanism, more
specifically, how the parameters of the slow K + currents regulate spike
generation. For understanding the impact of slow K + currents, we have
developed a framework to reduce a detailed conductance-based model
with slow K + currents to an adaptive threshold model [4]. We have
deduced a formula that links the slow K+ parameters to the parameters of
the reduced model. The formula was validated with the simulation of the
detailed model. This formula clarifies how I M and I AHP impact on spike
generation mechanism differently and the parameters of I M and I AHP
influence spike generation.
Acknowledgements: This study was supported by JSPS KAKENHI Grant
Number 24500372, 25870915, 25115728. We thank Shigeru Shinomoto
and Romain Brette for stimulating discussions.
References
1. Koch C: Biophysics of Computation Oxford, Oxford University Press 1999.
2. Prescott SA, Sejnowski TJ: Spike-rate coding and spike-time coding are
affected oppositely by different adaptation mechanisms. J Neurosci 2008,
28:13649-13661.
3. Deemyad T, Kroeger J, Chacron MJ: Sub- and suprathreshold adaptation
currents have opposite effects on frequency tuning. J Physiol 2012,
4. Kobayashi R, Tsubo Y, Shinomoto S: Made-to-order spiking neuron model
equipped with a multi-timescale adaptive threshold. Front Comput
Neurosci 2009, 3:9.
P126
Excitatory to inhibitory connectivity shaped by synaptic and
homeostatic plasticity
Claudia Clopath*, Jacopo Bono, Ulysse Klatzmann
Department of Bioengineering, Imperial College London, London, SW7 2AZ,
UK
E-mail: c.clopath@imperial.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P126
Recent experimental techniques allowed study of the relationship between
neurons stimulus-preference and connectivity. In particular, in the layer II/III
of primary visual cortex, it was shown that excitatory neurons with the same
orientation preference have a high probability of being bidirectionally
connected. However, the intracortical connectivity is only getting refined
after eye-opening. We have recently hypothesized that this process is a
result of experience-dependent plasticity, modelled by a Hebbian learning.
In contrast to excitatory neurons, parvalbumin-expressing (PV) inhibitory
cells are less input-specific: PV neurons receive excitatory inputs from
neurons with different orientation preferences. In this work, we investigate
the mechanism by which excitatory to inhibitory connections are formed
(how) and their potential function (why) in a small recurrent network. We
found that a model combining Hebbian learning with homeostatic plasticity,
which allows PV neurons to spike at a high rate (i.e reproducing the fast-
spiking intrinsic property of the cells), develops unspecific excitatory-to-
inhibitory connections (Figure 1). We then tested the role of inhibition by
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 90 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

Figure 1(abstract P126) Orientation preferences of excitatory and inhibitory neurons. A network of excitatory and inhibitory exponential integrate-
and-fire neurons with plastic feedforward inputs, where the recurrent connections from excitatory to excitatory and excitatory to inhibitory connections
are plastic under the voltage-triplet STDP rule. After learning, the excitatory neurons which have the same orientation preference have a high chance of
being bidirectional connected, as seen in experimentally. On the other hand, inhibitory neurons receive inputs from excitatory neurons with different
orientation preferences, consistent with recent experimental results. After learning, receptive fields of the excitatory neurons (black) and the inhibitory
neurons (blue). Note that the inhibitory neurons develop broader and more unspecific receptive fields than excitatory neurons

simulating our model with and without inhibition after learning
convergence. We found that inhibition ensures less fluctuation of the
synaptic weights over time, hence stabilizes the network. We therefore
propose that unspecific excitatory to PV connections can be a result of the
intrinsic homeostatic property of PV neurons, and can allow the network to
be more stable.
P127
Phase lead/lag due to degree inhomogeneity in complex oscillator
network with application to brain networks
Junhyeok Kim1, Joon-Young Moon2, Uncheol Lee2,3, George A Mashour2,3,
Seunghwan Kim1, Tae-Wook Ko4*
1
Nonlinear and Complex Systems Laboratory, Department of Physics, Pohang
University of Science and Technology, Pohang 790-784, Republic of Korea;
2
Department of Anesthesiology, University of Michigan Medical School, Ann
Arbor, MI 48109, USA; 3Center for Consciousness Science, University of
Michigan Medical School, Ann Arbor, MI 48109, USA; 4National Institute for
Mathematical Sciences, Daejeon 305-811, Republic of Korea
E-mail: twko@nims.re.kr
BMC Neuroscience 2015, 16(Suppl 1):P127
Brain anatomical connectivity is one of the main factors influencing
information flow among the brain areas [1] and phase lead/lag relationship
between oscillations of brain areas is known to be related to the
information flow [2,3]. In this study, we analyze the network effect on the
phases of coupled oscillators using Kuramoto model and obtain analytical
relationship between phase lead/lag and degrees of network nodes. We
also show robustness under various conditions, improving upon the result
of ref. [4]. Using the brain anatomical connectivity and the relationship, we
can explain the phase distribution across the brain. At first, we investigate
the relationship in the oscillator model on a scale-free network of which
degree distribution follows a power law [5]. Confirming the result of
previous study [4], the phases of higher degree nodes lag the phases of
lower degree nodes. Similar behaviors are observed also in random
network, where the degree distribution follows a Poisson distribution.
Using mean-field approximation, we analytically derive the relationship
between phases and node degrees as shown in Figure 1. With various
conditions of time delay and coupling strength, we also observe that this
phase lead/lag relationship between nodes is robust. Our exact
relationship can be well applied to human brain anatomical networks.
References
1. Stam CJ, van Straaten EC: Go with the flow: use of a directed phase lag
index (dPLI) to characterize patterns of phase relations in a large-scale
model of brain dynamics. Neuroimage 2012, 62:1415-1428.
2. Rabinovich MI, Afraimovich VS, Bick C, Varona P: Information flow
dynamics in the brain. Phys Life Rev 2012, 9:51-73.
3. Akam T, Kullmann DM: Oscillatory multiplexing of population codes for
selective communication in the mammalian brain. Nat Rev Neurosci 2014,
15:111-122.
4. Moon J-Y, Lee UC, Blain-Moraes S, Mashour GA: General relationship of
global topology, local dynamics, and information flow in simulated and
empirical brain networks. PLOS Comp Biol 2015, (to be published).
5. Ko T-W, Ermentrout GB: Partially locked states in coupled oscillators due
to inhomogeneous coupling. Phys Rev E 2008, 78:016203.
P128
Dynamics on global brain networks at the neuronal resolution
Masanori Shimono1,2*, Ryota Kobayashi3
1
Department of Physics, University of Indiana, Bloomington, IN, 47405, USA;
2
Harvard/MGH, 149 13th St, Charlestown, MA, 02129, USA; 3National Institute
of Informatics, 2-1-2 Hitotsubashi, Chiyoda, Tokyo 101-0003, Japan
E-mail: nori417@gmail.com
BMC Neuroscience 2015, 16(Suppl 1):P128
A key question in computational neuroscience is the way in which
microscopic components work together within the macroscopic brain
scale. We demonstrate a computational model simulating the whole brain
activity gathering neuronal components through columnar architectures.
We used a Multi-timescale adaptive threshold (MAT) model for the single
neuron model, which has proven to be one of the most accurate models
for reproducing the spike trains of a variety of cortical neurons in vitro [1].
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P127) The phase (black dots) of each oscillator and the calculated phases (red line) from analytic derivation. They match well
in locked region and red shaded area represents drifting region. The inset shows the degree for each node
The parameters used for expressing excitatory and inhibitory neurons were
adopted from it [2].
Before connecting through multiple brain regions, we tuned parameters
about background inputs and balance of synaptic intensities between
inhibitory cells and excitatory cells on local neuronal circuits [3,4]. The
background inputs were designed to represent both of driving inputs
from the Poisson neurons in subcortical nuclei and sustaining activities
within individual cortical regions.
A past study demonstrated computational simulation of the whole
mammalian brain including the thalamus by constructing a network
organization of the cortex using Diffusion Tensor Imaging (DTI) [5]. We used a
connectivity matrix provided from invasive tracing technique to sustain the
accuracy, and especially the directionality of connectivity matrix [6,7].
Furthermore, the cortical network used in our study includes weight of
connections [8]. We show that we can reconstruct the number of neurons
from the weight of connections, and can design connections between
neurons crossing different brain regions. The number of neurons at each brain
regions allowed us to integrate whole-brain network and neuronal dynamics
included in each brain region [7]. Computational modeling using non-invasive
brain images is also important to use in cases of human brain structure.
As summary, this computational model demonstrates a whole brain
dynamics at the resolution of the neuron level for the purpose of
illuminating what parameter will be potentially critical to change the
dynamics of the brain. From this computational simulation, we will show a
basis of understanding how optimally the brain structure is designed from
the generated dynamics, including robustness against damages on the brain.
Acknowledgements: This study was supported by a Grant-in-Aid for
JSPS Fellows for Research Abroad
References
1. Gerstner W, Naud R: How good are neuron models? Science 2009,
326:379-380.
2. Kobayashi R, Tsubo Y, Shinomoto S: Made-to-order spiking neuron model
equipped with a multi-timescale adaptive threshold. Frontiers in
computational neuroscience 2009, 3.
3. Potjans TC, Diesmann M: The Cell-Type Specific Cortical Microcircuit:
Relating Structure and Activity in a Full-Scale Spiking Network Model.
Cerebral Cortex 2012, bhs358.
4. Shimono M, Beggs JM: Functional clusters, hubs, and communities in the
cortical microconnectome. Cerebral Cortex 2014, bhu252.
Page 91 of 200
5. Izhikevich EM, Edelman GM: Large-scale model of mammalian
thalamocortical systems. Proceedings of the national academy of sciences
2008, 105(9):3593-3598.
6. Deco G, Jirsa V, McIntosh AR, Sporns O, Ktter R: Key role of coupling,
delay, and noise in resting brain fluctuations. Proceedings of the National
Academy of Sciences 2009, 106(25):10302-10307.
7. Shimono M: Non-uniformity of cell density and networks in the monkey
brain. Scientific reports 2013, 3.
8. Markov NT, Misery P, Falchier A, Lamy C, Vezoli J, Quilodran R, Gariel MA,
Giroud P, Ercsey-Ravasz M, Pilaz LJ, Huissoud C, Barone P, Dehay C,
Toroczkai Z, Van Essen DC, Kennedy H, Knoblauch K: Weight consistency
specifies regularities of macaque cortical networks. Cerebral Cortex 2011,
21(6):1254-1272.
P129
Subthreshold resonance in biophysically-based models of low- and
high-input conductance motoneurons
Vitor M Chaud*, Andr F Kohn
Biomedical Engineering Laboratory, Dept. of Telecommunication and Control
Engineering, University of So Paulo, So Paulo, SP, 05508-900, Brazil
E-mail: vitor@leb.usp.br
BMC Neuroscience 2015, 16(Suppl 1):P129
Resonance in the membrane potential enables a neuron to discriminate its
inputs on the basis of their frequency content, so that oscillatory inputs
near the resonant frequency produce the largest responses [1]. Recently,
Manuel et al. [2] have shown that spinal motoneurons (MNs) exhibit
membrane resonance in the frequency range of about 8-14 Hz. Modeling
studies of resonance in neural membrane usually use minimal or reduced
models, accounting only for the ionic currents sufficient to generate the
resonance and plausible to exist in the investigated neuron type. However,
in the present study we developed models of low- and a high-input
conductance MNs including most of the ionic currents known or
hypothesized to exist in mammalian MNs and exhibiting several of the
known physiological features. Therefore, instead of investigating resonance
as an isolated phenomenon we analyze it as consequence of interactions
between the passive properties and the ionic currents in a more realistic
fashion. The low-input conductance MN model did not show resonances,
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P129) Dependence of resonance of the high-input conductance MN model on the basal level of injected current. A. somatic
potential B. impedance amplitude C. impedance phase D. Quality (Q) factor E. resonance frequency (fr) F. total phase advance (FL)
whereas the high-input conductance MN model showed resonances
dependent on the membrane potential (Figure 1). These resonances were
mainly caused by the hyperpolarization-activated cationic (H) current.
However, for membrane potentials near the firing threshold (purple lines
in Figure 1), another resonance mainly affected by transient sodium and
delayed rectifier currents predominated and its resonance frequency was
correlated with the minimum firing rate caused by a step current injection.
Conclusion: These results contribute to the understanding of the
complex interactions between active and passive properties in MNs and
how these interactions can produce membrane potential resonance and
affect action potential generation. The relation we found between
subthreshold resonance and the minimum firing rate merits further
studies to better understand how firing rate may be affected by
membrane resonances.
References
1. Hutcheon B, Yarom Y: Resonance, oscillation and the intrinsic frequency
preferences of neurons. Trends in Neurosciences 2000, 23:216-222.
2. Manuel M, Meunier C, Donnet M, Zytnicki D: Resonant or not, two
amplification modes of proprioceptive inputs by persistent inward
currents in spinal motoneurons. J Neurosci 2007, 27:12977-12988.
P130
A super-resolution approach for receptive fields estimation of neuronal
ensembles
Daniela Pamplona1*, Gerrit Hilgen2, Bruno Cessac1, Evelyne Sernagor2,
Pierre Kornprobst1
1
INRIA Sophia Antipolis Mditerrane, Neuromathcomp, 2004 Route des
Lucioles, 06902 Valbonne, France; 2Institute of Neuroscience, Faculty of
Medical Sciences, Framlington Place, Newcastle upon Tyne, NE2 4HH
E-mail: daniela.pamplona@inria.fr
BMC Neuroscience 2015, 16(Suppl 1):P130
Page 92 of 200
The Spike Triggered Average (STA) is a classical technique to find a discrete
approximation of the Receptive Fields (RFs) of sensory neurons [1], a
required analysis in most experimental studies. One important parameter
of the STA is the spatial resolution of the estimation, corresponding to the
size of the blocks of the checkerboard stimulus images. In general, it is
experimentally fixed to reach a compromise: If too small, neuronal
responses might be too weak thus leading to RF with low Signal-to-Noise-
Ratio; on the contrary, if too large, small RF will be lost, or not described
with enough details, because of the coarse approximation. Other solutions
were proposed consisting in starting from a small block size and updating
it following the neuron response in a closed-loop to increase its response
[2-4]. However, these solutions were designed for single cells and cannot
be applied to simultaneous recordings of ensembles of neurons (since
each RF has its own size and preferred stimulus). To solve this problem, we
introduced a modified checkerboard stimulus where blocks are shifted
randomly in space at fixed time steps. This idea is inspired from super-
resolution techniques developed in image processing [4]. The main
interest is that the block size can be large, enabling strong responses,
while the resolution can be finer since it depends on the shift minimum
size. In [5] was shown that the STA remains an unbiased RF estimator and,
using simulated spike trains from an ensemble of Linear Nonlinear Poisson
cascade neurons, it was predicted that this approach improves RF
estimation over the neuron ensemble. Here, we test these predictions
experimentally on the RFs estimation of 8460 ganglion cells from two
mouse retinas, using recordings performed with a large scale high-density
multielectrode array. To illustrate the main interest of the approach, in
Figure 1 we show a representative example of STA for one neuron where
RFs have been obtained using the three following stimuli (all presented
during 15min, for one retina displayed at 10 Hz, for the other at 30 Hz): (A)
standard checkerboard stimulus with block size of 160m, (B) standard
checkerboard stimulus with block size of 40m, (C) checkerboard stimulus
with block size of 160m and arbitrary shifts of 40m in and y-directions.
Results show spatial resolution can be improved in case (C), while nothing
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P130) A, B, C: Representative example of STA of one neuron considering stimulus A, B) and C). D: Number of RFs mapped with
each stimulus
could be obtained in (B) by changing only the block size of the standard
stimulus. At the population level, plot (D) shows the number of the RFs
that could be recovered for each stimuli, using a decision criteria based of
the RFs value distribution. Most of the RFs were mapped with both
methods (A) and (C) (49.9%). However, the proposed case (C) allows to
recover 51% of the mapped RFs at a resolution of 40m, while in the
classical case (A), 41% of the RFs could be found at a resolution of only
160m. Thus, the method does improve the quality of the RF estimation
and the amount of successfully mapped RFs in neural ensembles.
Acknowledgements: This work was partially supported by the EC IP
project FP7-ICT-2011-9 no. 600847 (RENVISION).
References
1. Chichilnisky E J: A simple white noise analysis of neuronal light
responses. Network 2001, 12:199-213.
2. Foldiak P: Stimulus optimization in primary visual cortex. Neurocomputing
2001, 38-40:1217-1222.
3. Machens CK: Adaptive sampling by information maximization. Phys Rev
Lett 2002, 88(22):228104.
4. MacKay D: Information-based objective functions for active data
selection. Neural Comput 1992, 4:590-604.
5. Nasrollahi K, Moeslund T B: Super-resolution: a comprehensive survey.
Mach Vis Appl 2014, 25:1423-1468.
6. Cessac B, Kornprobst P: Shifting stimulus for faster receptive fields
estimation of ensembles of neurons. Cosyne Abstracts. Salt Lake City USA
2015.
P131
Nonparametric estimation of characteristics of the interspike interval
distribution
Ondrej Pokora1*, Lubomir Kostal2
1
Department of Mathematics and Statistics, Faculty of Science, Masaryk
University, Brno, Czech Republic; 2Institute of Physiology, Academy of
Sciences of the Czech Republic, Prague, Czech Republic
E-mail: pokora@math.muni.cz
BMC Neuroscience 2015, 16(Suppl 1):P131
We address the problem of non-parametric estimation of the probability
density function as a description of the probability distribution of
noncorrelated interspike intervals (ISI) in records of neuronal activity. We
also continue our previous effort [1,2] to propose alternative estimators of
the variability measures. Kernel density estimators are probably the most
frequently used non-parametric estimators of the probability distribution.
However, there are also other non-parametric approaches. We focus on non-
parametric methods based on a principle of extrema of the Fisher
information. Specifically, we focus on the maximum penalized likelihood
estimation of the probability density function proposed by Good and
Gaskins [3], which can be understood as a kernel estimator with a particular
kernel function [4]. Other non-parametric approach we would like to address
Page 93 of 200
is the spline interpolation proposed by Huber [5] which can uniquely
estimate the ISI distribution.
Acknowledgements: This work was supported by the Czech Science
Foundation (GACR) grants 15-06991S (Ondrej Pokora) and 15-08066S
(Lubomir Kostal).
References
1. Kostal L, Lansky P, Pokora O: Variability measures of positive random
variables. PLoS ONE 2011, 6:e21998.
2. Kostal L, Pokora O: Nonparametric Estimation of Information-Based
Measures of Statistical Dispersion. Entropy 2012, 14:1221-1233.
3. Good IJ, Gaskins RA: Nonparametric roughness penalties for probability
densities. Biometrika 1971, 58:255-277.
4. Eggermont PPB, LaRiccia VN: Maximum Penalized Likelihood Estimation:
Volume I: Density Estimation. Springer 2001.
5. Huber PJ: Fisher information and spline interpolation. Ann. Stat 1974,
2:1029-1033.
P132
The role of mechanosensory T cells for stimulus encoding in the local
bend network of the leech
Friederice Pirschel*, Oliver Kuehn, Jutta Kretzberg
Computational Neuroscience, University of Oldenburg, D-26111 Oldenburg,
Germany
E-mail: friederice.pirschel@uni-oldenburg.de
BMC Neuroscience 2015, 16(Suppl 1):P132
Accurate behavioral responses to sensory stimuli require reliable encoding
and processing of stimulus properties in the underlying neuronal network.
The relatively simple nervous system of the leech is able to react with
surprising precision when the skin is touched: With their local bend
response [1], leeches discriminate behaviorally between touch location
differences of 9 [2], corresponding to a distance of less than 1 mm. The
underlying network consists of one layer of sensory neurons, approximately
20 interneurons and a layer of motor neurons [1]. Studies investigating the
local bend mainly focused on one of the three mechanosensory cell types,
the pressure ("P) cells [1,2]. But decoding experiments revealed a
discrepancy between P cell activity and the behavioral performance [2].
Therefore, we investigated how the mechanosensory touch (T) cells,
respond to touch stimulus properties and influence responses of local
bend interneurons. Using a semi-intact preparation, we elicited the local
bend response with mechanical stimulation of the skin. Simultaneously,
we performed intracellular double recordings from T cells and
interneurons 157 or 159, which are involved in the local bend network
[3]. We found that T cells respond with characteristic bursts to the onset
of touch stimulation. The first T cell spikes are generated with an
extremely high temporal precision and short response latency for a broad
range of touch stimulus intensities. Local bend interneurons 157 and 159
get input from T cells and produce characteristic EPSPs with short
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 94 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

response latencies in response to tactile stimulation. In particular EPSPs of

cell 159 follow T cell bursts.
For quantitative data analysis, we used two maximum likelihood
approaches of stimulus estimation: The classification approach assigned
each response trial to the most probable of all possible stimuli. It provides
a measure, how well a certain response feature (e.g. spike count, latency,
EPSP amplitude) encodes a specific stimulus property (location, intensity,
duration). The pairwise discrimination approach compares responses to
two stimuli and quantifies minimal stimulus differences, which could be
detected based on the neuronal response features. These methods
revealed that the relative latency of two T cells leads to the best
estimation of stimulus locations (for tested intensities up to 50 mN). For
interneurons, graded response features (e.g. EPSP integral and amplitude)
allow good stimulus discrimination performance, even for small distances
of touch locations.
Additionally, we started building a computational model of the local bend
network. We tried to fit the parameters of the Izhikevich model [4] to
intracellularly recorded T cell responses. However, the depolarized resting
potential (compared to cortical neurons) of this invertebrate cell required
modification of the model. Moreover, we were not able to find a
parameter range reproducing the characteristic T-cell bursts in response to
stimulus onset. None of the tested parameter sets yielded stimulus-
dependent response latencies, which are essential for stimulus encoding.
Hence, the Izhikevich model is not well suited for modeling the response
properties of an invertebrate sensor cell, which are relevant for the
encoding of sensory stimuli.
Acknowledgements: Supported by the PhD program Neurosenses of Figure 1(abstract P133) Multivariate ISI, SPIKE and SPIKE-
the State of Lower Saxony. Synchronization profiles for 50 artificially generated spike trains.
References The dashed lines represent the respective expectation values for random
1. Kristan WB, Calabrese RL, Friesen WO: Neuronal control of leech behavior. Poisson spike trains
J Progress in Neurobiol 2005, 76:279-327.
2. Thomson EE, Kristan WB: Encoding and decoding touch location in the
leech CNS. J Neurosci 2006, 26:8009-8016.
3. Lockery SR, Kristan WB: Distributed processing of sensory information in methods to spike trains obtained from numerical simulations [7] as well as
the leech. II. Identification of interneurons contributing to the local experimental recordings [4].
bending reflex. J Neurosci 1990, 10:1816-1819. The methods are implemented in both the Matlab-based graphical user
4. Izhikevich EM: Simple model of spiking neurons. IEEE Trans Neural interface SPIKY [5] and the Python library PySpike.
Networks 2003, 14:1569-1572. Acknowledgements: This work was supported by the European
Commission through the Marie Curie Initial Training Network Neural
Engineering Transformative Technologies (NETT), project number 289146
P133 and through the European Joint Doctorate Complex oscillatory systems:
Time-resolved and parameter-free measures of spike train synchrony: Modeling and Analysis (COSMOS), project number 642563 (TK).
properties and applications References
Mario Mulansky*, Nebojsa Bozanic, Thomas Kreuz 1. Tiesinga P, Fellous J-M, Sejnowski TJ: Regulation of spike timing in visual
Institute for Complex Systems, CNR, Sesto Fiorentino, 50019, Italy cortical circuits. Nature Reviews Neuroscience 2008, 9:97-107.
E-mail: mario.mulansky@isc.cnr.it 2. Shlens J, Rieke F, Chichilnisky EJ: Synchronized firing in the retina. Current
BMC Neuroscience 2015, 16(Suppl 1):P133 opinion in Neurobiology 2008, 18(4):396-402.
3. Kreuz T, Haas JS, Morelli A, Abarbanel HDI, Politi A: Measuring spike train
The relevance of exact spike timings in neural coding was presumed synchrony. J Neurosci Methods 2007, 165:151.
since a long time and has now been experimentally established, see e.g. 4. Kreuz T, Chicharro D, Houghton C, Andrzejak RG, Mormann F: Monitoring
in [1,2]. A popular approach to the analysis of spike timings is to measure spike train synchrony. J Neurophysiol 2013, 109:1457.
the synchrony of spike trains. With the recent advancements of the 5. Kreuz T, Mulansky M, Bozanic N: SPIKY: A graphical user interface for
experimental techniques, it is now possible to simultaneously record the monitoring spike train synchrony. ArXiv 2015, 1410.6910v2 (Submitted to
activity of hundreds of neurons. The analysis of such collective responses JNeurophysiol).
requires new mathematical tools that are able to detect synchrony in 6. Mulansky M, Bozanic N, Sburlea A, Kreuz T: A guide to time-resolved and
groups of spike trains. Here, we present three methods to quantify spike parameter-free measures of spike train synchrony. Arxiv 2015, 1502.02027.
train synchrony that are applicable in such multivariate situations. All of (Submitted to IEEE).
these methods are parameter-free and time-resolved which makes them 7. Chen Y, Zhang H, Wang H, Chen Y: The role of coincidence-detector
easy to handle and able to detect temporal changes of synchrony. neurons in the reliability and precision of subthreshold signal detection
Specifically, we discuss the ISI-distance [3], the SPIKE-distance [4] and the in noise. PloS one 2013, 8(2):e56822.
very recently proposed SPIKE-Synchronization [5]. The ISI-distance is
based on the relative differences of interspike intervals, while the SPIKE-
distance uses exact spike timings. SPIKE-Synchronization can be P134
understood as a time-resolved, spike-wise coincidence detector. Figure 1 Detecting parallel bursts in in silico generated parallel spike train data
shows exemplarily the time-resolved profiles of all three methods for 50 Christian Braune*, Rudolf Kruse
artificially created spike trains. Institute for Knowledge Engineering and Language Processing, Otto von
We analyze the mathematical properties of all three measures and discuss Guericke University, Magdeburg, Germany
their advantages and disadvantages [6]. Specific focus lies on the statistical E-mail: christian.braune@ovgu.de
relevance of the obtained values compared to random spike trains. By BMC Neuroscience 2015, 16(Suppl 1):P134
calculating the expectation values for Poisson spike trains we are able to
provide an important point of reference for interpreting numerical and Introduction: Neurons process stimuli as joint groups [1]. With multi-
experimental results. Finally, we show exemplary applications of the electrode arrays being capable of recording hundreds of channels in
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
parallel the need for computational methods arises to efficiently find hints
for such groups in the recorded data. Enumerating all possible subsets of
neurons becomes quickly unfeasible if not virtually impossible to do.
Therefore, we developed methods for efficiently finding so-called
assemblies of synchronously firing neurons in spike train data [2,3].
However, these methods only consider nearly synchronous single
activations of neurons and ignore the non-stationary firing rates. It has
been shown that the bursting behavior of neurons is a different mode of
communication between neurons and has to be considered in the analysis
as well [4,5].
Method: Our method builds upon a previously released algorithm that
was intended to find synchronously activated neurons in parallel spike
train data. This method uses dynamically placed windows, centered on
each single spike, to detect episodes of increased synchrony among the
spike trains. By calculating the amount of overlap between a single spike
train and the complete set of spike trains new features can be generated.
These features allow to identify groups of neurons that show an increased
amount of synchronous activations compared to what would be expected
under the assumption of independence. By allowing the algorithm to
utilize information obtained from the burst detection process (e.g. [6-10])
we can efficiently and effectively identify those groups of neurons that
show increased synchronous bursting behavior.
Conclusion: Using artificially generated data we are able to test our
method on a multitude of data sets for which we actually know the true
assembly structure. The spike trains are generated in such a way, that their
statistical properties match those of in vitro recordings of embryonal cortical
slices, i.e. the inter-burst interval and intra-burst inter spike interval
distributions match. With this setup we test our algorithm on different
assembly numbers and sizes. We are then able to distinguish between non-
related and related neurons as well as to separate the related ones into
different assemblies.
References
1. Hebb DO: The Organization of Behaviour 1949, 62.
2. Borgelt C, Braune C: Prototype Construction for Clustering of Point
Processes based on Imprecise Synchrony. 8th conference of the European
Society for Fuzzy Logic and Technology (EUSFLAT-13) 2013.
3. Braune C, Borgelt C, Kruse R: Behavioral Clustering for Point Processes.
Advances in Intelligent Data Analysis XII. Springer 2013, 127-137.
4. Kaneoke Y, Vitek JL: Burst and oscillation as disparate neurophysiologic
properties: Method of detection. J Neurosci Meth 1996, 68:211-223.
5. Froemke RC, Dan Y: Spike-timing-dependent synaptic modification
induced by natural spike trains. Nature 2002, 416(March):433-438.
6. Bakkum DJ, Radivojevic M, Frey U, Franke F, Hierlemann A, Takahashi H:
Parameters for burst detection. Front Comput Neurosci 2013,
7(January):193.
7. Chen L, Deng Y, Luo W, Wang Z, Zeng S: Detection of bursts in neuronal
spike trains by the mean inter-spike interval method. Prog Nat Sci 2009,
19:229-235.
8. Chiappalone M, Novellino a, Vajda I, Vato a, Martinoia S, van Pelt J: Burst
detection algorithms for the analysis of spatio-temporal patterns in
cortical networks of neurons. Neurocomputing 2005, 65-66:653-662.
9. Cocatre-Zilgien JH, Delcomyn F: Identification of bursts in spike trains. J
Neurosci Methods 1992, 41:19-30.
10. Gourvitch B, Eggermont JJ: A nonparametric approach for detection of
bursts in spike trains. J Neurosci Methods 2007, 160:349-358.
P135
Minimal set of nodes to control the dynamics of biological neuronal
networks
Simachew Mengiste1,2, Ad Aertsen1, Arvind Kumar2*
1
Faculty of Biology and Bernstein Center Freiburg, University of Freiburg,
Freiburg im Breisgau, Germany; 2Computational Biology, School of Computer
Science and Communication, KTH, Stockholm, Sweden
BMC Neuroscience 2015, 16(Suppl 1):P135
Controlling the dynamics of biological neuronal network (BNNs) is of great
importance for both understanding the brain and for developing means to
control brain disease related aberrant activity. In the last decade, the
relationship among node properties, network structure and network
dynamics has become clearer and several neuron and network properties
could be used to control the network dynamics. Designing appropriate
Page 95 of 200
control inputs would require selective manipulation of the control variable
in each node. However, the large size of BNNs and scarcity of resources do
not allow control of all individual nodes by external means. Therefore,
there is a need to identify a much smaller number of influential nodes that
are sufficient to drive the activity dynamics of the entire BNN to a desired
state.
Here, we used the concept of structural controllability [1] to identify a small
set of influential nodes, control nodes, for synthetic and experimental
networks. The later includes large-scale connectivity of macaque cortical
regions [2], meso-scale connectome of a mouse brain [3] and the neuronal
connectivity of C. elegans [4]. We investigated the effect of random and
systematic edge pruning on the number of these control nodes. Moreover,
We showed that there is a power law or exponential dependence between
the number of control nodes and the average degree of many networks.
Finally, we used a simplified network of the basal ganglia to qualitatively
show that few control nodes can indeed drive the network dynamics to a
desired state. While every sub-network could be used as a control node or
sensor, each node is different in terms of the required stimulation pattern,
stimulation energy and time to reach the desired dynamical state.
Acknowledgements: Supported by Erasmus Mundus Joint Doctoral
programme EuroSPIN, the German Federal Ministry of Education and
Research (BMBF 01GQ0420 to BCCN Freiburg) and the Cluster of
Excellence BrainLinks-BrainTools funded by German Research Foundation
(DFG, grant number EXC 1086).
References
1. Lin C-T: IEEE Trans. Automat. Contrr 1974, 19:201-208.
2. Modha DS, Singh R: PNAS 2020, 107(30):13485-13490.
3. Oh SW, et al: Nature 2014, 508:207-214.
4. Watts DJ, Strogatz SH: Nature 1998, 393:440-442.
P137
A role of local VTA GABAergic neurons in mediating dopamine neuron
response to nicotine
Ekaterina Morozova1*, Maxym Myroshnychenko2, Marie Rooy3, Boris Gutkin3,4,
Christopher C Lapish5, Alexey Kuznetsov6
1
Department of Physics, Indiana University, Bloomington, IN, 47405, USA;
2
Program in Neuroscience, Indiana University, Bloomington, IN, 47405, USA;
3
Group of Neural Theory, ENS, Paris, 75005, France; 4National Research
University Higher School of Economics, Moscow, 101000, Russia; 5Addiction
Neuroscience Program, IUPUI, Indianapolis, IN, 46202, USA; 6Department of
Mathematics, IUPUI, Indianapolis, IN, 46202, USA
E-mail: emorozov@indiana.edu
BMC Neuroscience 2015, 16(Suppl 1):P137
The local circuitry of the tegmental area (VTA) consists primarily of
dopamine (DA) and GABA neurons. Interactions between DA and GABA
neurons are critical for regulating DA neuron activity, and thus DA efflux
throughout the brain. One striking example that demonstrates the
significance of local interactions between DA and GABA neurons is related
to nicotine reinforcement. Experimentally, it was shown that activation of
nicotinic acetylcholine (nAch) receptors on GABA neurons by Ach leads to
an increase in both firing and bursting of the DA neuron, while nicotine
produces an opposite effect [1]. In order to investigate the mechanism of
this GABA-mediated effect, we created a biologically plausible model of
local VTA microcircuitry. The model network consists of a population of
GABA neurons innervating one DA neuron. DA neuron dynamics are
described by a conductance based model; which includes intrinsic and
synaptic currents conducted by NMDA, AMPA, GABA, and nAch receptors.
GABA neurons that are described by Wang-Buszaki equations provide
inhibitory drive to the DA neuron. Excitatory inputs (Glu and Ach) to
DA and GABA neurons are modeled as Poisson-distributed spike trains.
Ach pulses act as synchronizing inputs to a population of GABA neurons,
due to a transient activation of GABA nAch receptors, while nicotine
persistently activates nAch receptor, causing an increase in firing
frequency of GABA neurons. Modeling revealed that synchronization of
GABA neurons by cholinergic input could provide a mechanism for the
elevation of DA neuron firing frequency and bursting. Synchronized GABA
inputs act via phasic disinhibition that promotes rebound spikes, but is
conditioned on the presence of depolarizing currents to the DA neuron.
The opposite effects of nicotine applied to the GABA neuron only
(decrease in firing and bursting of DA neuron), could be the result of
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 96 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

Figure 1(abstract P137) A) Quantification of Ach mediated firing of DA neuron (firing rate and %SWB) B) Nicotine-elicited modifications in
firing rate of DA neuron. Vertical black lines indicate nicotine injection. Horizontal dashed lines indicate the baseline firing rate. C) Example voltage
trace of DA neuron (nAch receptor on GABA neuron only)

desynchronization in population of GABA neurons, produced by tonic
activation of nAch receptor. A desynchronized GABA population provides
constant inhibition to the DA neuron that suppresses firing (see Figure 1).
These data highlight the important and powerful role local circuit
dynamics VTA. Furthermore, synchrony amongst GABA neurons seems to
be a critical intermediary that regulates DA neuron activity and, by
extension, DA release throughout the brain.
Reference
1. Tolu S, Eddine R, Marti F, David V, Graupner M, Pons S, Baudonnat M,
Husson M, Besson M, Reperant C, et al: Co-activation of VTA DA and
GABA neurons mediates nicotine reinforcement. Molecular Pshyciatry
2013, 18:382-393.
P138
Qualitatively different scenarios for co-activation of NMDA, AMPA and
GABA receptor currents on dopaminergic neuron
Denis Zakharov1,2*, Alexey Kuznetsov3
1
Department of Nonlinear Dynamics, Institute of Applied Physics, RAS,
Nizhny Novgorod, 603950, Russia; 2Laboratory of Nonlinear Physics, Nizhny
Novgorod State University, Nizhny Novgorod, 603022, Russia; 3Department of
Mathematical Sciences and Center for Mathematical Modeling and
Computational Sciences, IUPUI, Indianapolis, IN 46202, USA
E-mail: zakharov@neuron.appl.sci-nnov.ru
BMC Neuroscience 2015, 16(Suppl 1):P138
We introduced a reduced functional model that qualitatively describes
fundamental properties of the dopaminergic (DA) neuron. The neuron is
a low-frequency pacemaker, and requires special conditions to increase
the frequency of its firing [1]. In more detail, only N-methyl-D-aspartate
receptor (NMDAR) current can significantly elevate the frequency (more
than 5 fold in comparison with pacemaking). a-Amino-3-hydroxy-5-
methyl-4-isoxazolepropionic acid receptor (AMPAR) current is not able to
evoke such a high frequency. Instead, it suppresses firing at a
conductance much lower than NMDAR. g-Aminobutyric acid receptor
(GABAR) activation first decreases the firing frequency, and only then
suppresses firing. Thus, the latter property determines type 1 excitability
for the DA neuron, whereas the difference in responses to AMPA and
NMDA receptor activation make the neuron different from other well-
studied types. We focus this report on the co-activation of these
receptors. Co-activation of AMPA together with NMDA receptor may
obscure the high frequency firing and lead directly to its suppression.
There are contradictory experimental data in this case [2,3]. We show that
AMPAR can do both: further contribute to the frequency increase
produced by NMDAR, or impede the high-frequency firing. We show that
the distinction between these two cases is the level of the background
NMDAR activation. Co-activation of GABA together with NMDA receptor
may also lead directly to the suppression of high-frequency oscillations
(e.g. by shunting inhibition), or reduce the frequency first. Experiments
clearly show the latter [4]: the frequency can be reduced back to that in
control conditions when GABA and NMDA receptor currents are balanced.
This property allows us to recalibrate the model so that it retains type 1
excitability not only in control, but also with activated synaptic currents.
In general, activation of synaptic currents changes neural excitability, and
we investigate the dynamical mechanisms responsible for the transition.
As a result, we were able to construct a model of the DA neuron that
collects all the properties described above. The model is essential for
further research of behaviorally-relevant modulations of the DA neuron
activity, which involves simultaneous excitatory and inhibitory inputs to
the neuron.

Figure 1(abstract P138) Influence of the synaptic currents on the neuron activity. A and B: dependences of the frequency on the AMPAR and
NMDAR, GABAR and NMDAR conductances respectively. C: a voltage trace corresponding to the case of balanced GABAR and NMDAR currents
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Acknowledgements: This work was supported by the RFBR (project 14-
02-00916-a) and by the grant NIH-NIAAA R01AA022821.
References
1. Zakharov DG, Kuznetsov AS: A minimal model for a slow pacemaking
neuron. Chaos, Solitons & Fractals 2012, 45:640-644.
2. Deister CA, Teagarden MA, Wilson CJ, Paladini CA: An intrinsic neuronal
oscillator underlies dopaminergic neuron bursting. J Neurosci 2009,
29:15888-97.
3. Blythe SN, Wokosin D, Atherton JF, Bevan MD: Cellular Mechanisms
Underlying Burst Firing in Substantia Nigra Dopamine Neurons.
J Neurosci 2009, 29:15531-41.
4. Lobb CJ, Wilson CJ, Paladini CA: A Dynamic Role for GABA Receptors on
the Firing Pattern of Midbrain Dopaminergic Neurons. J Neurophysiology
2010, 104:403-413.
P139
Prefrontal-hippocampal theta coherence, sharp wave ripples, and
bursts of cortical unit activity underlie choices and encoding in the
radial arm maze
Maxym V Myroshnychenko1*, Christopher C Lapish2
1
Program in Neuroscience, Indiana University, Bloomington, IN 47401, USA;
2
Addiction Neuroscience Program, Indiana University-Purdue University,
Indianapolis, IN 46202, USA
E-mail: mmyros@gmail.com
BMC Neuroscience 2015, 16(Suppl 1):P139
The radial arm maze (RAM) is a foraging task that is often used to assess
executive function guided decision-making [1-3]. Optimal foraging
strategies on this task require the integration of executive and memory
systems, which include retrospective and prospective codes. To explore
the neural basis of decision-making during RAM performance, the current
Figure 1(abstract P139) Average hippocampal and prefrontal theta and SWR power centered on reward acquisition. During test choices (B, D),
PFC (D) activity resembles that in HC (B). On the other hand, during training (A, C), HC (A) shows the pattern of activity associated with reward sites, and
PFC (C) does not. Shaded regions, 95% confidence intervals
Page 97 of 200
study acquired multielectrode single-unit and local field potential (LFP)
recordings simultaneously in hippocampus (HC) and anterior cingulate
(ACC). Initially, rats were presented with four of the eight total maze arms
open at once ("training phase), and the rest of the arms were opened
up after the subjects collected the rewards and a one-minute delay was
completed ("test phase). Any arm re-entries were counted as errors, and
only trials with one test phase error or less were used in the analysis.
ACC unit firing was elevated at choice points and reduced at reward
points. Moreover, both ACC and the HC showed elevated theta power
shortly prior to the reward point and sharp wave ripples shortly after
reward acquisition on correct choices only, which is consistent with
findings in other choice tasks [4,5]. These observations held during test
phase (Figure 1, right panels) and were disrupted during the training
phase in ACC, but not the HC (Figure 1, left panels). It has been
suggested that HC sharp wave ripples contain episodes of replay of
visited locations and theta of locations ahead of the animal [4,5] -
information necessary for decision-making using both prospective and
retrospective codes. Theta and ripples are present in the HC during both
training and test phases, consistent with lesion evidence that hippocampus
is necessary for both prospective and retrospective strategies [1]. In ACC,
theta and ripples are only present during test phase and not during training
phase, which is in line the evidence showing that prefrontal cortex is only
necessary for prospective coding [1]. These task phase-dependent
observations may help explain how HC and ACC networks integrate
information related to prospective and retrospective codes.
References
1. Cook R, Brown M, Riley M: Flexible memory processing by rats: Use of
prospective and retrospective information in the radial maze.
J ExperPsych 1985, 11(3):453-69.
2. Lapish CC, Durstewitz D, Chandler LJ, Seamans JK: Successful choice
behavior is associated with distinct and coherent network states in
anterior cingulate cortex. PNAS 2008, 105(33):11963-8.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
3. Zhou W, Hohmann AG, Crystal JD: Rats answer an unexpected question
after incidental encoding. Current biology 2012, 22(12):1149-53.
4. Benchenane K, Peyrache A, Khamassi M, Tierney PL, Gioanni Y, Battaglia FP,
Wiener SI: Coherent theta oscillations and reorganization of spike timing.
Neuron 2010, 66(6):921-36.
5. Jadhav SP, Kemere C, German PW, Frank LM: Awake hippocampal sharp-
wave ripples support spatial memory. Science 2012, 336(6087):, 1454-8.
P140
Mode-locking behavior of Izhikevich neurons under periodic external
forcing
AmirAli Farokhniaee1*, Edward W Large1,2
1
Department of Physics, University of Connecticut, Storrs, CT, 06268, USA;
2
Department of Psychology, University of Connecticut, Storrs, CT, 06268, USA
E-mail: a.farok@uconn.edu
BMC Neuroscience 2015, 16(Suppl 1):P140
Many neurons in the auditory system of the brain must encode amplitude
variations of a periodic signal. These neurons under periodic stimulation
display rich dynamical states including mode-locking and chaotic
responses [1]. Periodic stimuli such as sinusoidal waves and amplitude
modulated (AM) sounds can lead to various forms of n:m mode-locked
states, similar to the mode-locking phenomenon in a LASER resonance
cavity. Obtaining Arnold tongues provides useful insight into the
organization of mode-locking behavior of neurons under periodic forcing.
In this study we obtained the regions of existence of various mode-
locked states on the frequency-amplitude plane, which are called Arnold
tongues, for Izhikevich neurons (see Figure 1). This study is based on the
model for neurons by Izhikevich (2003), which is a reduced model of a
Hodgkin-Huxley neuron [2]. This model is much simpler in terms of the
dimension of the coupled non-linear differential equations compared to
other existing models, but excellent for generating the complex spiking
patterns observed in real neurons [3]. Hence we can describe the
construction of harmonic and sub-harmonic responses in the early
processing stages of the auditory system, such as the auditory nerve and
cochlear nucleus.
Conclusion: Different mode-locked regions that are shown in the Arnold
tongues diagram are predictors of mode-locking of auditory system
Figure 1(abstract P140) The Arnold tongues diagram for a Class 1
Izhikevich neuron driven by an external sinusoidal forcing. This plot
represents the mode-locked regions as a function of the amplitude and
frequency of this sinusoidal forcing. Each colored region represents a
different phase-locked regime in terms of an integer ratio. Here, an n:m
ratio means the neuron produces n action potentials in response to
every m cycles of the stimulus. For example, for stimulus amplitudes and
frequencies corresponding to the yellow region, the neuron exhibits 3:1
mode-locking
Page 98 of 200
neurons to sound, which in turn predict the formation of harmonics and
sub-harmonics of the sound in the brain.
Acknowledgements: This work was supported by University of
Connecticut College of Liberal Arts and Sciences.
References
1. St-Hilaire Martin, Longtin Andre: Comparison of Coding Capabilities of
Type I and Type II Neurons. Journal of Computational Neuroscience 2004,
16:299-313.
2. Izhikevich MEugene : Dynamical Systems in Neurosceince: The geometry
of Excitability and Bursting. Cambridge, Massachusetts: The MIT Press
2007.
3. Kim Youngtae: Identification of Dynamical States in Stimulated Izhikevich
Neuron Models by Using a 0-1 Test. Journal of Korean Physical Society,
2010, 57(6):1363-1368.
P141
The OpenWorm Project: currently available resources and future plans
Padraig Gleeson1,2*, Matteo Cantarelli2, Michael Currie2, Jim Hokanson2,3,
Giovanni Idili2, Sergey Khayrulin2,4, Andrey Palyanov2,4, Balazs Szigeti2,5,
Stephen Larson2
1
Department of Neuroscience, Physiology and Pharmacology, University
College London, London, UK; 2OpenWorm Foundation, 3899 Nobel Drive,
Unit 1226, San Diego, CA 92122, USA; 3Department of Biomedical
Engineering, Duke University, Durham, NC, USA; 4Lab. of Complex Systems
Simulations, A.P. Ershov Institute of Informatics Systems, Siberian Branch of
the Russian Academy of Sciences, Novosibirsk, 630090, Russian Federation;
5
Neuroinformatics Doctoral Training Centre, University of Edinburgh,
Edinburgh, UK
E-mail: p.gleeson@ucl.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P141
The nematode C. elegans is historically one of the most studied model
organism in biology and the only one to have its full connectome
mapped. OpenWorm [1] is a project dedicated to recreating the C.
elegans nematode as a virtual organism in a computer. The project takes
an Open Science approach to development, relying on volunteer
contributions and making all code, data and documentation publicly
available at the time of production. The OpenWorm project has two long
term goals. The first is to functionally reproduce the behaviour of the
wild-type C. elegans in a variety of environmental contexts, to the extent
that the simulated behaviour is statistically indistinguishable from
recordings of real worms under analogous environmental conditions.
OpenWorms second goal is for the simulation to be a faithful biological
model for C. elegans. The achievement of these ambitious goals will
hopefully make OpenWorm a valuable software tool in C. elegans labs
worldwide. Scientists could make perturbations in-silico before beginning
the expensive and time-consuming work of conducting in-vivo
experiments. Conversely, having many scientist users will engender a
feedback process that will make development more data-driven, helping
to improve the biological realism of the OpenWorm model in the first
place.
The medium term aim of the project is to create infrastructure to enable
these longer term goals and tackle smaller but significant challenges in
whole organism modelling along the way. The first concrete milestone
the project has set for itself is an accurate simulation of the mechanical
activity of the worm including the motor system, with accurate
electrophysiology of the C. elegans muscle cells [2] to reproduce the
crawling gait. This model will be validated by comparing crawling
behaviour from experimental data. We aim to make this model accessible
through a web browser so that scientists and even interested members
of the public can explore and experiment with the model in 3D.
A number of activities are ongoing towards achieving this milestone:
Sibernetic: a Smoothed Particle Hydrodynamics based simulation engine
to simulate the environment and the body of the worm, with the ability
to simulate fluids of variable viscosities and contractible elastic matter
with impermeable membranes; Geppetto: a web based visualization and
simulation engine being developed to bring the output of the
OpenWorm project to the web for maximum accessibility, without the
need for local installation of the software required to run simulations of
the model; c302: a framework based on NeuroML for generating network
models of C. elegans incorporating known anatomical data which can
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
models for the neurons of varying levels of detail (point neurons to
multicompartmental); PyOpenWorm, a Python based API for accessing
the latest anatomical and physiological data on C. elegans, facilitating its
use in computational models, with data provenance built in from the
start; Movement Validation Framework: a suite of tools for analyzing
worm behaviour allowing extraction from videos (or from simulated
activity) of known motifs of worm movement.
All of the above applications are open source and can be reused for
modelling other species/brain regions. More details can be found on
http://www.openworm.org and https://github.com/openworm.
References
1. Szigeti B, Gleeson P, Vella M, Khayrulin S, Palyanov A, Hokanson J, Currie M,
Cantarelli M, Idili G, Larson S: OpenWorm: an open-science approach to
modelling Caenorhabditis elegans. Front. Comput. Neurosci 2014.
2. Boyle JH, Cohen N: Caenorhabditis elegans body wall muscles are simple
actuators. BioSystems 2008, 94(2008):170-181.
P142
Encoding of information using neural fingerprints
Jos Luis Carrillo-Medina1*, Roberto Latorre2
1
Departamento de Elctrica y Electrnica, Universidad de las Fuerzas
Armadas ESPE, Sangolqui, Ecuador; 2Grupo de Neurocomputacin Biolgica,
Dpto. de Ingeniera Informtica, Escuela Politcnica Superior, Universidad
Autnoma de Madrid, 28049 Madrid, Spain
E-mail: jlcarrillo@espe.edu
BMC Neuroscience 2015, 16(Suppl 1):P142
The existence of neural fingerprints associated to specific cell types or to
different processing states has been reported in widely different neural
systems (e.g. see [1-3]). Does the nervous system have the ability to
process information using these neural fingerprints? Model simulations
suggest that neural signatures characterizing the origin of specific signals
may play an important role for the fast and fine tuning of CPG rhythms [4],
being a possible mechanism to contextualize or discriminate neural
information in these circuits [5]. However, neural fingerprints can be a
mechanism to encode temporal information in other neural networks of
the nervous system.
In this work we investigate the ability of a simple network model to
implement a neural processing based on the emission and recognition of
neural fingerprints. We build two-dimensional networks of 50x50 binary
neurons that communicate by exchange of serial binary patterns. Each unit
is connected to eight other neurons using different network topologies. In
particular, we study the influence of small-world topologies with different
rewiring probability - ranging from regular to random networks - on the
self-organizing properties of the network. In addition to the eight
connection channels, another channel is used to introduce external stimuli
in the network. Each individual neuron has the ability to recognize
different neural fingerprints arriving thought its input channels in the form
of specific binary patterns, i.e. in the model a fingerprint consists of a
specific sequence of bits. When a fingerprint is recognized in the input, the
neuron emits the same fingerprint to all its neighbors with probability pr. If
no signature is recognized in a time step, the neuron emits a spontaneous
pattern with probability pe (pe << pr ). After emitting a pattern, neurons
have a refractory period during which neither emission nor recognition are
made.
The proposed simple network is able to generate complex collective
dynamics [6]. Without stimuli, the network evolves to a stationary state,
related to the emission of the spontaneous activity. This does not depend
on the network topology. When external stimuli are injected into the
network (both in series and in parallel), the different neural fingerprints
received through the external channels can propagate throughout the
whole ensemble and the collective dynamics drastically changes. In this
situation, localized patterns of activity related to the travelling neural
fingerprint arise in the network. The intraunit parameters and the
organization of connections between neurons tune the self-organizing
properties within the network and define the spatial organization of
the coexisting spatio-temporal patterns encoding the different stimuli.
These patterns can survive and compete for long periods after the end of
the stimulation, providing both short-term and long-term memory
mechanisms to the network. Random connections enhance the fingerprint
processing and potentiate long-term memory mechanisms. Conversely,
Page 99 of 200
regular patterns of connectivity promote short-term memory mechanisms
and increase the encoding/storage capacity. These results point out that
neural systems could process and encode information using neural
fingerprints.
Acknowledgements: Roberto Latorre was supported by MINECO
TIN2012-30883.
References
1. Szcs A, Pinto RD, Rabinovich MI, Abarbanel HD, Selverston AI: Synaptic
modulation of the interspike interval signatures of bursting pyloric
neurons. J Neurophysiol 2003, 89:1363-1377.
2. Klausberger T, Magill PJ, Mrton LF, Roberts JDB, Cobden PM, Buzski G,
Somogyi P: Brain-state- and cell-type-specific firing of hippocampal
interneurons in vivo. Nature 2003, 421:844-848.
3. Brochini L, Carelli PV, Pinto RD: Single Synapse Information Coding in
Intraburst Spike Patterns of Central Pattern Generator Motor Neurons.
J Neuroscience 2011, 31(34):12297-12306.
4. Latorre R, Rodrguez FB, Varona P: Effect of individual spiking activity on
rhythm generation of central pattern generators. Neurocomputing 2004,
56-60:535-540.
5. Latorre R, Rodrguez FB, Varona P: Neural signatures: multiple coding in
spiking-bursting cells. Biol Cybern 2006, 95:169-183.
6. Carrillo-Medina JL, Latorre R: Neural dynamics based on the recognition
of neural fingerprints. Front Comput Neurosci in press.
P143
Reconstructing dynamical models from optogenetic data
Sorinel A Oprisan1*, Patrick E Lynn2, Tamas Tompa3,4, Antonieta Lavin3
1
Department of Physics and Astronomy, College of Charleston, Charleston,
SC 29424, USA; 2Department of Computer Science, College of Charleston,
Charleston, SC 29424, USA; 3Department of Neuroscience, Medical University
of South Charleston, Charleston, SC 29424, USA; 4Faculty of Healthcare,
Department of Preventive Medicine, University of Miskolc, Miskolc, Hungary
E-mail: oprisans@cofc.edu
BMC Neuroscience 2015, 16(Suppl 1):P143
Optogenetics allows optical control of neuronal activity by using
genetically altered neural cells and optical tools. Briefly, optogenetics uses
a photosensitive element that, upon absorption of light, produces some
change in the activity of the cells. Although the technique evolved, it
involves inserting a light-sensitive channel from green algae, called
channelrhodopsin-2 (ChR2), into neurons [1]. By precise spatial and
temporal delivery of light pulses we can identify the local interconnections
among neurons and investigate their dynamical response under different
conditions.
Data were recorded with a custom made optrode made of a recording
pipette glued to optic fiber inserted in the medial prefrontal cortex (PFC)
of male PV-Cre mice injected with virus suspension AV2/5.EF1a.DIO.hChR2
(H134R)-EYFP.WPRE.hGH [2]. A 473 nm laser generated light stimulation
and stable single unit recordings were monitored before filtering to
record field potentials (0.1-100 Hz). Optical stimulation consisted of 10 ms
pulses followed by 15 ms pause (40 Hz). A train of 10 pulses that lasted
250 ms was applied every 4 seconds and local filed potentials were
recorded with a sampling rate of 10 KHz (Figure 1A). We used delay-
embedding method to reconstruct the phase space attractor [3]. We
found that the minimum dimension that unfolds the attractors is three
and the delay time is about 3600 data points (Figure 1B). Based on the
phase space reconstruction, we were able to extract a low-dimensional
mathematical model that describes the dynamics of the system. Although
every single neuron in the mPFC is described by a large number of
independent variables such as ionic channel activation/inactivation
variables, the local neural network activated by light pulses can be
modeled with only three variables. We hypothesize these three global
variables could be the activity of excitatory, inhibitory interneurons, and
light-sensitive neurons.
Acknowledgements: Oprisan acknowledges IOS CAREER grant 1054914
from the National Science Foundation
References
1. Boyden SEdward, Zhang Feng, Bamberg Ernst, Nagel Georg, Deisseroth Karl:
Millisecond-timescale, genetically targeted optical control of neural
activity. Nature Neuroscience 2005, 8:1263-1268.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P143) A. Average local filed potential (LFP) recordings that show the response of mPFC to 250 ms light stimuli. B. Phase
space reconstruction of LFP using delayed embedding method in a three dimensional space and a delay time of 3600 data points (360 ms)
2. Dilgen EJonathan, Saggu Shalini, Naselaris Thomas, Lavin Antonieta:
Optogenetically evoked gamma oscillations are disturbed by cocaine
administration. Frontiers in cellular neuroscience 2013, 7:1-9.
3. Takens Floris: Detecting Strange Attractors in Turbulence in Dynamical
Systems and Turbulence. Warwick 1980, Springer Lecture Notes in
Mathematics 1981, 898:366.
P144
Computational model of medial temporal lobe epilepsy
Sora Ahn1, Sangbeom Jun1, Hyang Woon Lee2, Seungjun Lee1*
1
Department of Electronics Engineering, Ewha Womans University, Seoul,
120-750, Korea; 2Department of Neurology, Ewha Womans University, Seoul,
120-750, Korea
E-mail: slee@ewha.ac.kr
BMC Neuroscience 2015, 16(Suppl 1):P144
Temporal lobe epilepsy represents a high proportion of whole epilepsy
patients. Medial temporal lobe epilepsy (MTLE) is generated from internal
structures like hippocampus, and patients with MTLE are poorly controlled
by antiepileptic drugs [1]. Recently, deep brain stimulation (DBS) that is to
control seizure activity by stimulating epileptic zone is receiving attention
as a new treatment of epilepsy. However, the exact mechanisms are still
unclear and the current method is being developed relying on clinical
experiences. Consequently, researches for etiology of disease along with
seizure suppress mechanisms by electrical stimulation are very significant.
These studies would be best progressed with complementary cooperation
Figure 1(abstract P144) Recording data (left) and simulation result (right) of seizure-like events in entorhinal cortex
Page 100 of 200
between in-vitro and in-vivo experiments, and computer simulations using
a computational model.
In this paper, we propose a hippocampal network model which portrays
seizure-like events (SLEs) recorded in in-vitro experiments. The model is
composed of excitatory and inhibitory neurons interconnected following
the well-known synaptic pathway to form a small world network [2]. Each
neuron is descripted by Izhikevichs model [3] and synaptic current is
calculated based on conductance of a receptor. Short-term and long-term
plasticity are also applied to every synapse [4]. SLEs induced by 4-AP are
divided into three regions according to time-frequency features. The first
region is transition to ictal region by excitatory GABAergic drive [5], the
second region is tonic firing region by synchronization due to recurrent
excitation between principle neurons [6], and the last region is clonic
bursting and termination region by GABA-mediated inhibitory mechanisms
[1]. Proposed model faithfully reproduces these phenomena by controlling
synaptic input gain.
The effectiveness of the model is confirmed by comparing the simulation
results with experimental data which were recorded in rat hippocampal
slice in 4-AP bath application using micro-electrode array (MEA). Below
Figure 1 are time domain signals generated from computer model and
recorded in in-vitro measurement, respectively.
References
1. Avoli M, DAntuono M, Louvel J, Khling R, Biagini G, Pumain R,
DArcangelo G, Tancredi V: Network and pharmacological mechanisms
leading to epileptiform synchronization in the limbic system in vitro.
Prog Neurobiol 2002, 68:167-207.
2. Netoff TI, Clewley R, Arno S, Keck T, White JA: Epilepsy in small-world
networks. J Neurosci 2004, 24(37):8075-8083.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
3. Izhikevich EM: Simple model of spiking neurons. IEEE Trans Neural Netw
2003, 14(6):1569-1572.
4. Izhikevich EM, Gally JA, Edelman GM: Spike-timing dynamics of neuronal
groups. Cereb Cortex 2004, 14(8):933-944.
5. Staley KJ, Soldo BL, Proctor WR: Ionic mechanisms of neuronal excitation
by inhibitory GABAA receptors. Science 1995, 269:977-981.
6. Ziburkus J, Cressman JR, Barreto E, Schiff SJ: Interneuron and pyramidal
cell interplay during in vitro seizure-like events. J Neurophysiol 2006,
95:3948-3954.
P145
PyMICE - a Python library for analysis of mice behaviour
Jakub M Kowalski1*, Alicja Pucian1, Zofia Mijakowska2, Maria Nalberczak2,
Kasia Radwaska2, Szymon ski1
1
Department of Neurophysiology, Nencki Institute of Experimental Biology,
Warsaw, 02-093, Poland; 2Department of Molecular and Cellular Neurobiology,
Nencki Institute of Experimental Biology, Warsaw, 02-093, Poland
E-mail: j.kowalski@nencki.gov
BMC Neuroscience 2015, 16(Suppl 1):P145
Manual analysis of abundant behavioral data produced by automated
systems for long-time monitoring of a group of animals is extremely
inefficient and error prone. Some systems (like IntelliCage) are shipped
with software enough for basic analysis of the data, however lacking
flexibility for more advanced analyses.
To facilitate research reproducibility, same analysis of same data should
always yield same results. One (possibly the best) way to achieve such
robustness is to have the process automated (due to the IT slogan let the
computer do the work), therefore reducing number of possible human
errors. The other advantage of such approach is that modern computers are
both faster and more precise than humans when dealing with numbers.
An automated analysis workflow can be created easily if there is a possibility
of a convenient access to data. That is the purpose of development of
PyMICE library by our laboratory. The library provides its user with an object
oriented application programming interface (API) and a data abstraction
layer - therefore shifting ones focus from the form the data is provided to
the data itself. A simple analysis can be performed in just a few lines of
readable source code (see Figure 1 for an example). Moreover, the library
comes with auxiliary tools supporting development of analysis workflows.
Some of them facilitate data validation while other are dedicated for
workflow configuration.
Acknowledgements: Sponsored by Symfonia NCN grant: UMO-2013/08/
W/NZ4/00691.
P146
A biologically based neural network model for decision making
Azadeh Hassanejad Nazir1*, Hans Liljenstrom1,2*
1
Department of Energy and Technology, SLU, Uppsala, SE-75007, Sweden;
2
Agora for Biosystems, Sigtuna, SE-19322, Sweden
E-mail: azadehhn@kth.se; hans.liljenstrom@slu.se
BMC Neuroscience 2015, 16(Suppl 1):P146
We present a neural network model, describing an adaptive decision
making process (DM), under varying internal and external contexts [1].
Figure 1(abstract P145) A six-line example of data analysis with
PyMICE library. Firstly the library is loaded, secondly data are read from
file data.zip. In the third line a list of all visits of mouse named Jerry is
obtained. Then the first nose poke (if any) from every visit is selected
and then - the side of the nose poked door. Finally numbers of left and
right doors nose poked as first are counted
Page 101 of 200
The model includes the three most crucial structures in both emotional
and rational aspects of DM: amygdala, orbitofrontal cortex (OFC) and
lateral prefrontal cortex (LPFC) [2-4]. Neural activities in these structures
represent emotional attitude, expectancy value, and rules towards the
options (Figure 1). The DM is modeled at a level of mesoscopic
neurodynamics, using biologically inspired neural networks [5]. Rational
and emotional associations are encoded with cell assembly oscillations in
all three structures, determining the value of an option, V(opt), as a
product of the number of activated cells and the mean frequency and
amplitude of their oscillations (Figure 2). A decision is based on the
competition among stored patterns, using cosine similarity of the
frequency vectors f . The option with highest value will win the
competition in each system:
V(opt)|sopt |.fopt , Aopt  = |sopt |.fopt , Aopt , opt = 1, ..., n
The emotional system is goal directed on the predicted value of the
options. Emotional memory of various options are formed and stored in
the amygdala. A prediction signal is generated in OFC and an emotional
response is measured as a level of satisfaction.
LPFC is a pivotal structure in the rational analysis and is based on habitual
DM. Declarative and procedural memories, as the main components of this
system are activated by external stimuli. The bidirectional interactions
between maintained rules in procedural memory, rational attitudes
towards different options, and stored factual information in declarative
memory, lead to the selection of an option as a rational decision.
The integration of emotional and rational activities results in a final decision.
When an action is executed as a result of a decision its experienced value is
compared with the expected one, and stored in memory. Depending on the
sign and the magnitude of the prediction error (expected value - real value),
the stored emotional and rational attitude might be updated.
The experience of our decisions/choices are learnt and may influence future
decisions. For any particular input signal, the final decision could shift,
depending on internal and external context. Large delayed rewards have a
lower value, compared to small immediate rewards. This fact can be
included with the help of a hyperbolic discounting function, which models
the exponential reduction of rewards in terms of time.
References
1. Doyay K: Modulators of decision making. Nature Neuroscience 2008,
11:410-416.
2. Whalen PJ, Phelps EA: The Human Amygdala The Guilford Press 2009.
3. Zald D, Rauch S: The Orbitofrontal Cortex Oxford University Press 2006.
4. Dixon ML, Christoff KE: The lateral prefrontal cortex and complex value-
based learning and decision making. Neurosci Biobehav Rev 2014,
45C:9-18.
5. Liljenstrm H: Modeling the dynamics of olfactory cortex using simplified
network units and realistic architecture. International Journal of Neural
Systems 1991, 2:l-15.
Figure 1(abstract P146) Interactions of the main neural structures
in the decision making process: Amygdala, orbitofrontal cortex
(OFC) and lateral prefrontal cortex (LPFC). Inputs from sensory
modalities and from rational and emotional feedback
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 2(abstract P146) Activity of a cell assembly representing one of possible options. Red and green traces show the activity of the inhibitory
neurons and the blue one the activity of all excitatory neurons
P147
Numerical characterization of noisy fluctuations in two different types
of stochastic differential equation models of neural signaling
Tiina Manninen1*, Jukka Intosalmi2, Keijo Ruohonen3, Marja-Leena Linne1
1
Department of Signal Processing, Tampere University of Technology,
Tampere, Finland; 2Department of Computer Science, Aalto University,
Espoo, Finland; 3Department of Mathematics, Tampere University of
Technology, Tampere, Finland
E-mail: tiina.manninen@tut.fi
BMC Neuroscience 2015, 16(Suppl 1):P147
Stochastic modeling plays an essential role in the study of noise and
random fluctuations in biochemical signaling of neural systems. Here, we
study numerically noisy fluctuations produced by two different stochastic
differential equation models, the chemical Langevin equation (CLE) model
[1] and the rate constant stochastic differential equation (rcSDE) model [2],
using a biologically realistic neuronal protein kinase C activation pathway
[3] as a case study. We compare the CLE and rcSDE models by studying
the noise power levels in different system volumes and with different
model parameters. In this context, we also assess the problem of negative
concentrations by computing the average number of state vectors
containing negative concentrations within a single simulation run. Based
on the information obtained from the examination of noise power levels,
we then choose appropriate model parameters for the rcSDE model and
the corresponding system volume for the CLE model. Using these
parameters and the volume, we finally simulate the models and compare
the results using multiresolution analysis.
In the simulations we present here, the rcSDE model is capable of producing
the same level of the average power of noise with a notably smaller amount
of negative concentrations than the CLE model. This difference is
considerable and it suggests that in certain simulation settings, it might be
reasonable to use rcSDE model as an approximate simulation approach if it
is feasible to use the slightly different interpretation of the nature of noise.
A more detailed multiresolution analysis reveals that the resulting noise
processes differ at the higher frequencies. In many studies these differences
at high frequencies are not of interest as neural biochemical systems often
produce relatively stronger fluctuations at low frequencies. In some cases,
however, noise might be important for the system function and, in such
Page 102 of 200
situations, it is crucial to make sure that the noise is modeled in an
appropriate way. Whether the appropriate way is the CLE, the rcSDE model,
or some other model is dictated by the application. Different models are
based on different premises and it is modelers task to pick up an appropriate
model. We conclude that introduction of techniques from the field of signal
processing may help assessing the role of noise in neural systems and what
type of model to use. This study shows how the orthogonal wavelet
representation or similar multiscale decompositions can be successfully
applied to the study of noise in biochemical processes of neural systems.
Acknowledgements: This work was supported by the Academy of Finland,
application numbers 106030, 124615, 126556, and 129657 (Finnish
Programme for Centres of Excellence in Research 2006-2011), as well as the
Emil Aaltonen Foundation, Otto A. Malm Foundation, and Tampere Doctoral
Programme in Information Science and Engineering (TISE). The work also
benefited from Tampere Center for Scientific Computing and the grid
computing solution provided by Techila Technologies Ltd.
References
1. Gillespie DT: The chemical Langevin equation. J Chem Phys 2000,
113(1):297-306.
2. Manninen T, Linne M-L, Ruohonen K: Developing Ito stochastic differential
equation models for neuronal signal transduction pathways. Comput Biol
Chem 2006, 30(4):280-291.
3. Bhalla US, and Iyengar R: Emergent properties of networks of biological
signaling pathways. Science 1999, 283(5400):381-387.
P148
Extending computational models of astrocyte-neuron interactions with
biochemical mechanisms on the postsynaptic terminal
Aura Saudargien1, Tiina Manninen2, Riikka Havela2, Marja-Leena Linne2*
1
Department of Informatics, Vytautas Magnus University, Kaunas, Lithuania;
2
Department of Signal Processing, Tampere University of Technology,
Tampere, Finland
E-mail: marja-leena.linne@tut.fi
BMC Neuroscience 2015, 16(Suppl 1):P148
There is increasing evidence that astrocytes not only interact with each
other but also with adjacent neurons in the neural circuitry in a variety of
brain areas. Many biophysical and biochemical mechanisms have been
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
proposed to explain these interactions in vitro. Based on the experimental
literature, the mechanisms involved seem to depend not only on the
developmental stage of an animal but also on the brain area, neural
circuitry, as well as on the experimental technique used to characterize
the phenomena. Using biophysically and biochemically detailed
compartmental models, we are interested in understanding how these
interactions between neurons and astrocytes may regulate information
processing and, notably, different forms of synaptic plasticity [1]. In this
study, we designed a new model of the so-called tripartite synapse. The
tripartite synapse is a concept of synaptic physiology in which there are
three parts of a synapse: the presynaptic and postsynaptic terminals, and
an astrocyte in between them [2]. Many previous studies in vitro have
shown that gliotransmitters like glutamate are released from astrocytes
into the synaptic cleft following activation of presynaptic or postsynaptic
terminals, or activation of the astrocyte itself [3]. However, none of the
previous modeling studies incorporate all features necessary to
understand the complex interactions in full detail. We extend beyond
these previously published fundamental, yet relatively simplistic
representations (see, e.g., [4]). We describe the postsynaptic terminal
using a two-compartmental approach, in addition to the presynaptic
terminal and the astrocyte. We specifically address the postsynaptic
mechanisms and their role in activating astrocytic calcium signaling and
subsequent gliotransmission. By introducing realistic model components
on the postsynaptic terminal, such as voltage-dependent receptor
channels and the G-protein activated signaling cascades, we observe
stimulus-dependent changes in astrocyte calcium oscillations leading to
activation of not only astrocytes, but also of adjacent neurons. The long-
term goal of our work is to develop detailed models of astrocyte-neuron
interactions for different brain areas which allow testing experimentally
evoked hypothesis, as well as some controversies in the field (see, e.g.,
[5]). These models will open up new avenues to assess the effects of
numerous mechanisms of astrocytes on the dynamics of local neuronal
networks. Only through systematic integration of in vitro and in silico
work will we be able to understand how astrocytes may contribute to
brain information processing and plasticity.
Acknowledgements: The research leading to these results has received
funding from the European Union Seventh Framework Programme (FP7/
2007-2013) under grant agreement n 604102 (HBP). This work was also
supported by the Academy of Finland, application number 126556, as
well as the Emil Aaltonen Foundation and Otto A. Malm Foundation for
T.M. Tampere University of Technology Graduate School is acknowledged
for supporting R.H.
References
1. Paixo S., Klein R: Neuron-astrocyte communication and synaptic
plasticity. Curr Opin Neurobiol 2010, 20:466-473.
2. Araque A, Parpura V, Sanzgiri RP, Haydon PG: Tripartite synapses: glia, the
unacknowledged partner. Trends Neurosci 1999, 22(5):208-215.
3. Parpura V, Haydon P: Physiological astrocytic calcium levels stimulate
glutamate release to modulate adjacent neurons. Proc Natl Acad Sci USA
2000, 97:8629-8634.
4. Volman V, Ben-Jakob E, Levine H: The astrocyte as gatekeeper of synaptic
information transfer. Neural Comput 2007, 19:303-326.
5. Volterra A, Liaudet N, Savtchouk I: Astrocyte Ca2+ signalling: an
unexpected complexity. Nature Rev Neurosci 2014, 15:327-335.
P149
Information-theoretic analysis of a dynamic release site using a two-
channel model of depression
Mehrdad Salmasi1,2*, Martin Stemmler3,4, Stefan Glasauer1,2,3,5, Alex Loebel3,4
1
Graduate School of Systemic Neurosciences, Ludwig-Maximilian University,
Munich, Germany; 2German Center for Vertigo and Balance Disorders,
Ludwig-Maximilian University, Munich, Germany; 3Bernstein Center for
Computational Neuroscience, Munich, Germany; 4Department of Biology II,
Ludwig-Maximilian University, Munich, Germany; 5Department of Neurology,
Ludwig-Maximilian University, Munich, Germany
E-mail: mehrdad.salmasi@lrz.uni-muenchen.de
BMC Neuroscience 2015, 16(Suppl 1):P149
Synapses are dynamic communication channels between neurons as their
rates of information transfer depend on past history. While information
theory has been used to study the information efficacy of synapses [1-3],
Page 103 of 200
Figure 1(abstract P149)
the effect of synaptic dynamics, including short-term depression and
facilitation, on the information rate is not yet fully understood.
To reduce the complexity of the problem, we confine ourselves here to a
single release site of the synapse. This allows us to analytically calculate
the information transfer at the release site for a simple model of synaptic
depression which is based on binary channels. The input of the model is
a spike train, modeled by an independent identically distributed process
X = {Xi } i=1 , where each Xi has a Bernoulli distribution with P(Xi = 0) =
a. The models output is a process Y = {Yi }
i=1 , such that if there is a
release at time, then Yi = 1 and otherwise Yi = 0. We model the short
term depression by two binary asymmetric channels that represent the
possible states of the release site: the recovered state, when no release
occurred in the previous time step (Figure 1a), and the used state,
following vesicle release (Figure 1b). In particular, we assume that the
release probability is reduced following a release, that is p2 p1 and 1 -
q 2 1 - q 1.
Each individual channel in Figure 1 will have a mutual information rate, either
r1or r2. As Xiis Bernoulli-distributed, ri = h(pi + q
i ) h(pi ) h(q
i)
for i = 1,2, where h() is the entropy of a Bernoulli random variable and
x = 1 x . We prove that the mutual information rate of the release site with
depression is a linear summation of the information rates of these two channels.
The mutual information rate I(X;Y) between the input process X nd the output
p2 + q 2
process Y, is I(X;Y) = r1 + (1 - )r2 where = .
(p1 + p2 ) + (q1 + q2 )
The closed form expression of the mutual information rate allows us to
study the effect of depression analytically. Through simulations we show
that for a range of parameters, depression improves the rate of
information transfer at the release site. We also show that when the level
of depression is increased (i.e., with smaller p2 and larger q2), the release
sites information capacity is reached at lower input spike rates.
Therefore, the optimal spike rate of the presynaptic neuron has a reverse
relationship with the depression level of its release site. This means that
synaptic depression can save energy while maintaining information rate.
The two-channel model of release site is a building block for the
construction of more precise models of synaptic transmission. These
advanced models will enable us to evaluate and study the synaptic
information rates analytically.
Acknowledgement: This work was supported by the BMBF grant
01EO1401 (German Center for Vertigo and Balance Disorders)
References
1. London M, Schreibman A, Husser M, Larkum ME, Segev I: The information
efficacy of a synapse. Nature Neuroscience 2002, 5(4):332-340.
2. Fuhrmann G, Segev I, Markram H, Tsodyks M: Coding of temporal
information by activity-dependent synapses. Journal of Neurophysiology
2002, 87(1):140-148.
3. Goldman MS: Enhancement of information transmission efficiency by
synaptic failures. Neural Computation 2004, 16(6):1137-1162.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
P150
Interplay of intrinsic and network heterogeneity in strongly recurrent
spiking networks
Cheng Ly
Department of Statistical Sciences and Operations Research, Virginia
Commonwealth University, Richmond, VA 23284, USA
E-mail: CLy@vcu.edu
BMC Neuroscience 2015, 16(Suppl 1):P150
Heterogeneity has recently gained a lot of attention and it is becoming
more apparent that it is a crucial feature in neural processing [1-5]. Despite
its importance, this realistic physiological feature has traditionally been
neglected in theoretical studies of cortical neural networks. A common
reason is that mean-field descriptions of noisy cortical networks are high
dimensional and generally intractable. Although heterogeneous spiking
neural networks have recently been studied theoretically [5-8], there is still
a lot unknown. In particular, combining network heterogeneity [9] and
intrinsic heterogeneity [10] have yet to be considered simultaneously
despite the fact that both are known to exist and likely have significant
roles in neural network dynamics.
To this end, we study a recurrently coupled spiking network of leaky
integrate-and-fire (LIF) neurons consisting of excitatory and inhibitory
neurons. The intrinsic heterogeneity is modeled by varying the voltage
threshold for spiking [5], and the network heterogeneity is modeled by
different conductance strengths (partially motivated by recent results [11],
both excitatory and inhibitory conductances are scaled so each neuron has
a different level of balanced input). Unsurprisingly, we find that when
either intrinsic or network heterogeneity is increased, the response
heterogeneity also increases (i.e., the range of the average firing rate of the
excitatory neurons also increases). However, for a fixed level of both forms
of heterogeneity, the network robustly exhibits a wide range of response
heterogeneity that strongly depends on the relationship between intrinsic
and network heterogeneity. This coupled network is difficult to analyze
because it is stochastic, heterogeneous, and high dimensional with alpha
function synapses and colored external noisy input. With combination of
Monte Carlo simulations and augmented mean-field theory based partially
on methods in [12-15], we provide analytic explanations to account for the
observed phenomena. Our work gives insight for how these two forms of
heterogeneity interact in a generic recurrent spiking network that may be
applicable to many areas of the cortex.
References
1. Shamir M, Sompolinsky H: Implications of neuronal diversity on
population coding. Neural Computation 2006, 18:1951-1986.
2. Chelaru MI, Dragoi V: Efficient coding in heterogeneous neuronal
populations. Proceedings of the National Academy of Sciences 2008,
105:16344-16349.
3. Padmanabhan K, Urban NN: Intrinsic biophysical diversity decorrelates
neuronal firing while increasing information content. Nature neuroscience
2010, 13:1276-1282.
4. Tripathy SJ, Padmanabhan K, Gerkin RC, Urban NN: Intermediate intrinsic
diversity enhances neural population coding. Proceedings of the National
Academy of Sciences 2013, 110:8248-8253.
5. Mejias JF, Longtin A: Optimal heterogeneity for coding in spiking neural
networks. Physical Review Letters 2012, 108:228102.
6. Ly C: Dynamics of Coupled Noisy Neural Oscillators with Heterogeneous
Phase Resetting Curves. SIAM Journal on Applied Dynamical Systems 2014,
14:1733-1755.
7. Nicola W, Campbell SA: Mean-field models for heterogeneous networks
of two-dimensional integrate and fire neurons. Frontiers in computational
neuroscience 2013, 7:184.
8. Zhou P, Burton SD, Urban NN, Ermentrout GB: Impact of neuronal
heterogeneity on correlated colored noise-induced synchronization.
Frontiers in computational neuroscience 2013, 7:113.
9. Perin R, Berger TK, Markram H: A synaptic organizing principle for cortical
neuronal groups. Proceedings of the National Academy of Sciences 2011,
108:5419-5424.
10. Marder E: Variability, compensation, and modulation in neurons and circuits.
Proceedings of the National Academy of Sciences 2011, 108:15542-15548.
11. Xue M, Atallah BV, Scanziani M: Equalizing excitation-inhibition ratios
across visual cortical neurons. Nature 2014, 511:596-600.
Page 104 of 200
12. Moreno-Bote R, Parga N: Auto- and crosscorrelograms for the spike
response of leaky integrate-and-fire neurons with slow synapses. Physical
Review Letters 2006, 96:028101.
13. Nesse WH, Borisyuk A, Bressloff PC: Fluctuation-driven rhythmogenesis in
an excitatory neuronal network with slow adaptation. Journal of
Computational Neuroscience 2008, 25:317-333.
14. Ly C: A Principled Dimension-Reduction Method for the Population
Density Approach to Modeling Networks of Neurons with Synaptic
Dynamics. Neural Computation 2013, 25:2682-2708.
15. Nicola W, Ly C, Campbell SA: One-dimensional Population Density
Approaches to Recurrently Coupled Networks of Neurons with Noise [http://
arxiv.org/abs/1411.2273], Submitted.
P151
Parallelizing large networks using NEURON-Python
Alexandra H Seidenstein1*, Robert A McDougal2, Michael L Hines2,
William W Lytton3,4
1
Dept of Chemical & Biomolecular Engineering, New York University, NY,
USA; 2Dept of Neurobiology, Yale University New Haven CT, USA; 3Kings
County Hospital Center, Brooklyn, NY, USA; 4Dept. of Physiology &
Pharmacology, SUNY Downstate Medical Center, Brooklyn, NY, USA
E-mail: ahs@neurosim.downstate.edu
BMC Neuroscience 2015, 16(Suppl 1):P151
Research on brain organization has become increasingly dependent on large
multiscale neuronal network simulations. Here we describe current usage of
the NEURON simulator with MPI (message passing interface) in Python for
simulation of networks in the high performance computing (HPC) parallel
computing environment [1-4]. The mixed ordinary differential equation
(ODE) and event-driven nature of neuronal simulations offers advantages for
parallelization, by allowing each node to work independently for a period
equivalent to the minimal synaptic delay before exchanging queue
information with other nodes, obviating the need to exchange information
at every time step.
NEURONs ParallelContext allows access to a few of the important general
collective MPI calls, as well as calls adapted from prior usages from the
LINDA package, now reimplemented under MPI. From Python, a NEURON
ParallelContext is created using pc = h.ParallelContext(), where h
provides access to NEURON simulation objects after from neuron
import h, ParallelContext permits the periodic transfer of spike
information via queue exchanges.
Pseudo-random streams must be consistent regardless of numbers of
nodes in order to set connectivity, delays, and weights that are not fully
defined from experimental studies. These streams are kept consistent
regardless of number of nodes being used and therefore allows for the
simulations to be identical. In order to create this, randomizers are
established for particular purposes using NEURONs h.Random().
Random123().[5] The key to reproducibility is to define each randomizer
according to 1. a particular usage, 2. a particular cell (based on a global
identifier or gid), 3. a particular run based on a run identifier runid: e.
g., after for r in randomizers: r.Random123_globalindex
(runid, randomdel.Random123(id32(randomdel), self.
gid, 0) where id32() provides a 32-bit hash for a name: def id32
(obj):return hash(obj)&0xffffffff.
Data saving must be initially managed at node of origin and then
combined across nodes, to be accessible for analysis or viewing. Given
that file saving may occur incrementally during simulation from different
nodes on different local filesystems, file management becomes important.
There are several ways to handle data saving, the benefits and
applicability of which will be presented. Spike recordings are created as
vectors on a per-node basis, later consolidated. Other state variables may
also be saved at the same time, or may be recreated later utilizing a re-
run of individual cells with identical stimulation using NEURONs
PatternStim.
We note that ParallelContext in NEURON permits the development of
hybrid networks using various types of cells: event-driven cells, integrate-
and-fire cells, multicompartment cells, as well as complex cells with
calculation of internal chemical milieu. Load balancing in the hybrid
circumstance is a crucial issue, particularly when some cells are
computationally large due to inclusion of reaction-diffusion mechanisms
to develop multiscale models from molecule to network.
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 105 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

References
1. Carnevale NT, Hines ML: The NEURON Book Cambridge University Press,
New York 2006.
2. Hines ML, Carnevale NT: NEURON: a tool for neuroscientists. Neuroscientist
2001, 7:123-135.
3. Hines ML, Carnevale NT: Translating network models to parallel hardware
in neuron. J Neurosci Methods 2008, 169:425-455.
4. Migliore M, Cannia C, Lytton WW, Hines ML: Parallel network simulations
with NEURON. J. Computational Neuroscience 2006, 6:119-129.
5. Salmon JK, Moraes MA, Dror RO, Shaw DE: Parallel random numbers: as
easy as 1,2,3. Proceedings of 2011 International Conference for High
Performance Computing, Networking, Storage and Analysis, November 12-18,
2011, Seattle, Washington 2011.
P152
Motor cortex neurons: from experiment to model via evolutionary
algorithms
Samuel A Neymotin1*, Benjamin A Suter2, Michele Migliore3,
Salvador Dura-Bernal1, Gordon MG Shepherd1, William W Lytton1,4
1
Department Physiology & Pharmacology, SUNY Downstate, Brooklyn, NY,
11203, USA; 2Department Physiology, Northwestern University, Chicago, IL,
60611, USA; 3Institute of Biophysics, National Research Council, Palermo, Italy;
4
Department of Neurology, Kings County Hospital Center, Brooklyn, NY,
11203, USA
E-mail: samn@neurosim.downstate.edu
BMC Neuroscience 2015, 16(Suppl 1):P152
There are several types of neurons in primary motor cortex, with
differences in ion channels, dynamics, morphology, as well as differences
in synaptic input localization and axonal projections. To constrain
computer models of cortical neurons, it is necessary to record from these
different types of neurons. Model parameters can then be optimized, to
minimize the difference between the model and experimental data.
We have performed experiments recording from 2 classes of layer 5
pyramidal neurons in mouse motor cortex: corticospinal (SPI) and
corticostriatal (STR). These neurons exhibit differences in synaptic inputs,
axonal projections, and intrinsic membrane properties. Somatic whole-cell
recordings were performed in both classes and current steps were injected
(-0.2 to 0.6 nanoamperes) to produce different dynamical behaviors,
including hyperpolarization-induced sag, alterations in firing rate and spike
timing. We have developed multiple computer models of the different
neuronal classes, each with varying complexity, and all constrained by
experimental data. Two optimization methods were used: 1. evolutionary
algorithms, 2. NEURONs principal axis (praxis) error minimization algorithm.
We optimized model parameters with multiple fitness functions, first
independently, and then in different combinations. The fitness functions
quantified differences between experimental and simulated data, for firing
frequency, resting membrane potential (RMP), action potential shape,
interspike interval voltage trajectory, spike timing, hyperpolarization-induced
sag, and voltage responses to onset/offset of current injections.
We assessed optimized performance of different classes of neuronal
compartmental models, with varying levels of morphological complexity,
across the fitness functions. The models included a 5-compartment
pyramidal neuron (with soma, basal dendrite, and three apical dendritic
segments), and a ~904-compartment pyramidal neuron with detailed
dendritic and axonal geometry reconstructed from our experiments. All
model neurons included the same set of ion channels: INa and IKdr for
action potential generation; IKa for rapid repolarization following an action
potential; IKm and IKd for afterhyperpolarization; Ih for resonance, sag,
excitability, and contribution to RMP; calcium channels (L, N, T-type) for
bursting, excitability modulation, and contribution to backpropagating
action potentials; calcium-activated potassium channels for regulating
excitability after depolarization-induced calcium influx.
We performed optimization in two phases: 1. optimize passive properties
and compartmental geometry; 2. optimize maximal conductance and
kinetics of each ion channel. The distribution of ion channels were
constrained by patterns from experimental literature. For example, Ih
increased exponentially with distance from the soma, while somatic and
axonal compartments had higher densities of Na and K channels to allow
action potential generation. This also reduced free parameters for
optimization. Across the different models, the firing frequency, sag, and
RMP had nearly perfect fits to the experimental data, while using a spike
timing fitness function reduced spike time errors to within 1-10 ms. Due to
the high dimensionality of parameter space, different subsets of fitness
functions tended to improve independently, while others were correlated.
We assessed which parameter values and ion channels had the largest
impact on each fitness function using a post-optimization sensitivity
analysis. Our methods were effective in optimizing both SPI and STR
neurons and will likely generalize to other cell types.
Acknowledgements: Research supported by NIH grant U01EB017695.
P153
Large-scale M1 microcircuit model with plastic input connections from
biological PMd neurons used for prosthetic arm control
Salvador Dura-Bernal1*, Cliff C Kerr2, Samuel A Neymotin1, Bejamin A Suter3,
Gordon MG Shepherd3, Joseph T Francis1, William W Lytton1
1
Department of Physiology and Pharmacology, SUNY Downstate, Brooklyn,
NY 11203, USA; 2School of Physics, University of Sydney, Sydney, NSW,
Australia; 3Department Physiology, Northwestern University, Chicago, IL,
60611, USA
E-mail: salvadordura@gmail.com
BMC Neuroscience 2015, 16(Suppl 1):P153
We have developed a hybrid model of motor control for a brain-machine
interface that is based on a large-scale model of primary motor cortex (M1)
based on several mammalian studies (Figure 1). The M1 model consisted
of 10,000 spiking Izhikevich neurons of four types: regular-firing and
bursting pyramidal neurons, and fast-spiking and low-threshold-spiking
interneurons. Within the M1 population, cell proportions, locations,
connectivity and delays were drawn primarily from mouse experimental
data. Properties were based on cell body cortical depth (distance from pia
to white matter). Synapses included four different receptors: AMPA, NMDA,
GABAA and GABAB. The model exhibited realistic physiological properties,
including firing rates and local field potential spectra.
Pyramidal tract-projecting neurons in layer 5B were connected to a
descending spinal cord neural population, which provided excitation to the
muscles of a realistic virtual musculoskeletal arm. Proprioceptive feedback
from the arm was encoded in an ascending spinal cord population which
then projected to M1 layer 2/3. The virtual arm movements were also
followed by a robotic arm [1].
An additional population, which reproduced the spiking patterns
recorded from 92 neurons of macaque dorsal premotor cortex (PMd)
Figure 1(abstract P153) Model of primary motor cortex showing
inputs from PMd biological neurons, and closed-loop interaction
with virtual musculoskeletal arm. Cell types are shown with different
colors. The output connections of a single L5A neuron are shown to
illustrate connectivity
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
during a center-out reaching task, was connected to M1 layer 5A
providing a modulatory input [2].
The network was trained to drive the virtual arm to reach multiple targets,
by combining arm exploratory movements with reinforcement learning and
spike-timing dependent plasticity (STDP). Synaptic plasticity occurred
between multiple populations, including between the PMd inputs and layer
5A neurons. Tuning of learning metaparameters was achieved by employing
parallel evolutionary algorithms in a high-performance computing cluster.
This work moves towards a new generation of neuroprosthetic systems
where biological brain circuits interact directly with biomimetic cortical
models, and employ co-adaptation and learning to achieve a functional task.
Such a multiscale approach, ranging from the cellular to the behavioral level,
will provide deeper insights into brain dynamics and have applications for
the diagnosis and restoration of brain disorders.
References
1. Dura-Bernal S, Chadderdon GL, Neymotin SA, Francis JT, Lytton WW:
Towards a real-time interface between a biometic model of
sensorimotor cortex and a robotic arm. Pattern Recognition Letters 2014,
36(15):204-212.
2. Lee G, Matsunaga A, Dura-Bernal S, Zhang W, Lytton WW, Francis JT,
Fortes JAB: Towards real-time communication between in vivo
neurophysiological data sources and simulator-based brain biomimetic
models. Journal of Computational Surgery 2014, 3:12.
P154
A lower bound on the number of mechanisms for discriminating fourth
and higher order spatial correlations
John WG Seamons*, Marconi S Barbosa, Anton Bubna-Litic, Ted Maddess
Eccles Institute for Neuroscience, John Curtin School of Medical Research,
ANU, Canberra, ACT 0200, Australia
E-mail: john.seamons@anu.edu.au
BMC Neuroscience 2015, 16(Suppl 1):P154
The human visual system must employ mechanisms to minimize
informational redundancy whilst maximizing dynamic range and
maintaining that which is behaviorally relevant [1,2]. Previous research has
concentrated on two-point correlation properties, as captured by spatial
frequency and orientation tuning. There has been less research into higher-
order correlations although they may inform us about cortical functioning
[3]. Isotrigon textures can be used to probe the sensitivity of the human
visual system. The obvious structure in isotrigons is exclusively due to 4th
and higher-order spatial correlations [4]. Thus, in order to discriminate
isotrigons from noise, it is necessary to identify higher-order structure.
Although artificially generated, the same structural features that give
isotrigons salience also create salience in natural images [2].
Factor analysis can be used to infer the number of underlying independent
neurological mechanisms which govern isotrigon discrimination. In this
study, mean performance functions were calculated for two subjects using
ten new isotrigons (VnL2) (Figure 1A). Two forms of factor analysis
identified 3 principal factors (Figure 1B) [5]. Previous studies support that
the number of mechanisms is less than 10 [6], and more likely 2-4 [7,8].
Such mechanisms may represent some combination of recursive or
rectifying processes. Simple models of cortical processing, based on
recursion, can discriminate isotrigons [9]. The formation of recursively
applied products is physiologically plausible and can occur via dendritic
back-propagation or dendritic spiking [10].
References
1. Barlow H: Redundancy reduction revisited. Network 2001, 12(3):241-253.
2. Tkacik G, Prentice JS, Victor JD, Balasubramanian V: Local statistics in
natural scenes predict the saliency of synthetic textures. Proc Natl Acad
Sci U S A 2010, 107(42):18149-18154.
3. Victor JD: Isodipole Textures: A Window on Cortical Mechanisms of Form
Processing. Early Vision and Beyond MIT Press: Papathomas TVC, C.; Gorea,
A; Kowler, E 1995, 99-107.
4. Maddess T, Nagai Y, James AC, Ankiewcz A: Binary and ternary textures
containing higher-order spatial correlations. Vision Res 2004,
44(11):1093-1113.
5. Seamons JWG, Bubna-Litic A, Barbosa MS, Maddess T: A Lower bound on
the Number of Mechanisms for Discriminating Fourth and Higher Order
Spatial Correlations. Vision Res 2015.
Page 106 of 200
6. Taylor RR, Maddess T, Nagai Y: Spatial biases and computational constraints
on the encoding of complex local image structure. J Vis 2008, 8(7):19 11-13.
7. Maddess T, Nagai Y: Discriminating isotrigon textures [corrected]. Vision
Res 2001, 41(28):3837-3860.
8. Maddess T, Nagai Y, Victor JD, Taylor RR: Multilevel isotrigon textures.
J Opt Soc Am A Opt Image Sci Vis 2007, 24(2):278-293.
9. Nagai Y, Taylor RR, Loh YW, Maddess T: Discrimination of complex form
by simple oscillator networks. Network 2009, 20(4):233-252.
10. Hausser M, Spruston N, Stuart GJ: Diversity and dynamics of dendritic
signaling. Science 2000, 290(5492):739-744.
P155
Spatiotemporal brain network analysis of healthy humans based on
magnetoencephalography and functional MRI in the resting state
Margaret Y Mahan1,2*, Arthur C Leuthold2,3, Apostolos P Georgopoulos1,2,3
1
Graduate Program in Biomedical Informatics and Computational Biology,
University of Minnesota, Minneapolis, MN 55455, USA; 2Brain Sciences Center,
VA Health Care System, Minneapolis, MN 55417, USA; 3Department of
Neuroscience, University of Minnesota, Minneapolis, MN 55455, USA
E-mail: mahan027@umn.edu
BMC Neuroscience 2015, 16(Suppl 1):P155
Modern graph theory provides powerful quantitative tools and measures for
the analysis of complex networks. The network measures are critical for
creating metrics that allow for the comparison of connectivity patterns and
neural interactions across subjects and brain measurements. In addition,
these measures provide a framework for the characterization of network
structure by identifying local contributions of individual nodes and
connections, as well as the networks global capacity to integrate
information. Since neural communication is the basis of brain function,
assessing the dynamic status of the brain as a network is important.
However, a major challenge is how to reliably measure and assess the
network, and evaluate its utility. For that purpose, we used resting-state
magnetoencephalography (MEG) and functional magnetic resonance
imaging (fMRI) recordings from 65 cognitively healthy women to reliably
assess the brain as a dynamic network.
MEG time series were prewhitened using ARIMA(50,1,1) and fMRI time
series were prewhitened using ARIMA(15,1,1) to yield practically white
noise innovations. For MEG, nodes were defined to be the individual
sensors (n = 248); for fMRI, a modified functional template [1] was used
to define nodes equaling those used for MEG in order to compare
networks across the two modalities. Connectivity matrices were then
calculated using pairwise Pearson correlations between the processed (1)
whole time series, and (2) windowed time series (fMRI = 6 s window,
MEG = 50 ms window) of all nodes. Weighted graphs of each of these
connectivity matrices were evaluated using the following metrics: (a) local
connectivity (strength, diversity), (b) global connectivity (mean correlation
coefficient), (c) local network properties (degree, clustering coefficient,
local efficiency, betweenness centrality), and (d) global network
properties (smallworldness, centrality, global efficiency, path length,
assortativity, transivity).
For the whole series, this analysis yielded MEG and fMRI static networks
for each subject, characterized by the set of the values obtained for the
parameters above. The effect of age was assessed by performing a
multivariate regression where the network parameters were the
dependent variables and age was the independent variable. Through an
iterative, stepwise procedure those parameters that varied systematically
with age were detected and retained. This analysis was performed for
both the MEG and fMRI data, and the resulting MEG- and fMRI-based
network parameters were compared. Similarly, for the windowed series,
the same analyses were performed per time window to yield a dynamic
network variation and the variation/stability of network structure over
time was assessed. This provided an additional temporal network
measure for evaluating the effect of age on brain network structure. We
believe that this approach is at the heart of assessing brain networks at
varying time scales, at different states in health and disease, and during
performance of different tasks.
Reference
1. Power JD, Cohen AL, Nelson SM, Wig GS, Barnes KA, Church JA, Vogel AC,
Laumann TO, Miezin FM, Schlaggar BL, Petersen SE: Functional network
organization of the human brain. Neuron 2011, 72:665-678.
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 107 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

Figure 1(abstract P154) A: Mean performance functions for all subjects by texture type, presented separately according to subject and texture
size. For All16 (16x16 textures) and All32 (32x32) error bars are SE for n = 6 subjects. Glider shapes are shown in the bottom panel. B: Communalities for
5 different factor models. As the number of factors grows, the profile of bars becomes flatter indicating that the models progressively account for the
data in a more balanced way. After nf = 3, the improvement in the reconstruction is marginal

P156
Pattern recognition of Hodgkin-Huxley equations by auto-regressive
Laguerre Volterra network
Kunling Geng1,2*, Vasilis Z Marmarelis1,2
1
Biomedical Engineering Department, University of Southern California, Los
Angeles, CA 90089, USA; 2Biomedical Simulations Resource, Los Angeles, CA
90089, USA
E-mail: kgeng@usc.edu
BMC Neuroscience 2015, 16(Suppl 1):P156
A nonparametric, data-driven nonlinear auto-regressive Volterra (NARV) [1]
model has been successfully applied for capturing the dynamics in the
generation of action potentials, which is classically modeled by Hodgkin-
Huxley (H-H) equations. However, the compactness still need to be
improved for further interpretations. Therefore, we propose a novel Auto-
regressive Sparse Laguerre Volterra Network (ASLVN) model (shown in
Figure 1A), which is developed from traditional Laguerre Volterra Network
(LVN) and principal dynamic mode (PDM) framework [2].
We adopt stochastic global optimization algorithm Simulated Annealing
[3] to train the ASLVN instead of Back-propagation method [2] to avoid
local minima and convergence problems. We also use lasso regularization
[4] to enhance the spasity of the network and prune redundant branches
for parsimony. The prediction results are shown in Fig.1B, it can be seen
that the exogenous output z(1) represents the subthreshold dynamics in
phase III, and the autoregressive output z(2) dominates in the spike shape
in phase I, and the cross term output z(x) helps to maintain the refractory
period by cancelling the effect of z(1) in phase II and we also observe that
refractory inhibition effect decays after initiation of AP, which explains the
absolute refractory period and relative refractory period in physiology.
References
1. Eikenberry SE, Marmarelis VZ: A nonlinear autoregressive Volterra model
of the Hodgkin-Huxley equations. Journal of computational neuroscience
2013, 34(1):163-183.
2. Marmarelis VZ: Nonlinear dynamic modeling of physiological systems.
John Wiley & Sons 2004, 10.
3. Kirkpatrick S: Optimization by simmulated annealing. science 1983,
220(4598):671-680.
4. Tibshirani R: Regression shrinkage and selection via the lasso. Journal of
the Royal Statistical Society Series B (Methodological) 1996, 267-288.
P157
Neural coding of monaural and binaural intensity at low stimulus
frequencies
Zbynek Bures1,2*, Petr Marsalek3,4
1
Dept. of Electric Engineering and Computer Science, College of
Polytechnics, Jihlava, 586 01, Czech Republic; 2Inst. of Experimental Medicine,
Academy of Sciences of the Czech Republic, Prague, 142 20, Czech Republic;
3
Dept. of Pathological Physiology, First Faculty of Medicine, Charles
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P156) A Structure of ASLVN for modeling H-H equations, where the input x(n) is the randomly injected current and the
output y*(n) is the membrane potential. B The predictions results, z(1) represents the exogenous output, z(2) represents the autoregressive output and
z(x) represents the cross term output
University in Prague, 128 53, Czech Republic; 4Czech Technical University in
Prague, Zikova 1903/ 4, 166 36, Czech Republic
E-mail: Zbynek.Bure@svspj.cz
BMC Neuroscience 2015, 16(Suppl 1):P157
At low sound frequencies, spikes in the auditory nerve (AN) are phase-
locked to the stimulus waveform. For sinusoidal stimuli, spike occurrences
are restricted to the positive half-waves of the sound cycles, which
constraints possible values of spike rates and their variability. As the spike
rates are used for representation of sound intensity, the just noticeable
differences (JNDs) of sound level and of inter-aural level difference (ILD)
will be influenced. Due to the lack of appropriate AN data, we explore
the topic using a computational model [1].
Three possible types of spike trains are considered: a synchronous, an
asynchronous spike train with variable intensity, and a periodically
Page 108 of 200
restarting (with sound period) asynchronous process. Generally, the
synchronous process leads to several times lower JNDs than the other
processes. The JNDs depend on further parameters: longer counting
windows or higher number of converging fibers result in smaller JNDs,
larger spike timing jitter (random displacement of spikes from their ideal
position) slightly increases the JNDs. Notably, the dependence of JNDs on
mean spike rate is in contradiction to psychophysical observations that
report monotonic decrease of JND of intensity with increasing sound
pressure level, the so called near-miss to Webers law [2]. Even with ideal
phase-lock and no jitter, the simulated JND of the synchronous algorithm
has a bell shape, the other two algorithms give monotonically increasing
function (Figure 1A, colored lines). Recruitment of more AN fibers with
rising intensity overcomes this discrepancy; a simulation of this
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P157) Just-noticeable differences of sound level depending on spike rate (A) and sound frequency (B)
mechanism yields results well matching the psychophysical results (Figure 1A,
black line).
Another discrepancy of the modeling results with psychophysics is the
simulated JND increasing with frequency, given a constant spike rate
(Figure 1B), while psychophysical JNDs decrease with frequency. However,
if we assume that maximum spike rate is in some way limited by the
frequency (e.g., at most one spike can occur in each sound cycle) and
that a given intensity range (e.g., 30 dB) is always mapped to the
available range of spike rates (e.g., between zero and sound frequency),
then the slope of the rate-intensity function increases and the JND
decreases with frequency. The results show that intensity coding has
specific properties at low sound frequencies, deserving more detailed
electrophysiological studies than what is available in experimental
literature today.
Acknowledgements: Funded by Institutional Support for Long-term
Development of Research Organizations (PRVOUK) no. P 24 at the Charles
University in Prague to P. M.
References
1. Bures Z, Marsalek P: On the precision of neural computation with inter-
aural level differences in the lateral superior olive. Brain Res 2013,
1536:16-26.
2. Ward LM, Davidson KP: Where the action is: Weber fractions as a
function of sound pressure at low frequencies. J Acoust Soc Am 1993,
94(5):2587-2594.
P158
Voltage sensitive currents and information processing by single neurons
Eduard Kuriscak1*, Zdenek Wunsch1,2, Petr Marsalek2,3*
1
Institute of Physiology, First Faculty of Medicine, Charles University in
Prague, 128 00, Czech Republic; 2Inst. of Pathological Physiology, First Faculty
of Medicine, Charles University in Prague, 128 53, Czech Republic; 3Czech
Technical University in Prague, Zikova 1903/ 4, 166 36, Czech Republic
E-mail: Eduard.Kuriscak@lf1.cuni.cz; Petr.Marsalek@lf1.cuni.cz
BMC Neuroscience 2015, 16(Suppl 1):P158
This is a computational study using data and models. We study selected
neurons in neural circuits, which have different precision of spike timing
under reproducible experimental conditions. Voltage sensitive currents
govern foremost the spike generation at given times. Yet the effects of
voltage sensitive currents are not limited to the axonal hillock, other
neuronal structures are also involved. We compare previously described
neural computations using models ranked according to level of detail
and complexity.
Single neurons propagate neural signal by following steps. The first is the
synaptic processing, realized by function of pre- and post- synaptic
membrane. Next is signal recoding and encoding and information
processing, which is realized by dendrites. As next step follows binary
decision spike generation with yes-or-no logical value. This occurs at the
axonal hillock. Further signal recoding and encoding is computed by
axonal propagation. Finally there is again synaptic processing at the
synapse to the next neuron.
Page 109 of 200
In past, some salient features amongst these steps were implemented in
biophysically detailed neuronal simulations of anatomically reconstructed
neurons. This was achieved via brute-force numerical solution of many
parallel instances of cable equation solved by numerical solver of partial
differential equations named GENESIS [1]. The experimental observation
of voltage dependent ion currents (active currents) in dendrites has been
a milestone confirming high complexity and high information throughput
in single neurons [2].
We use computational description of the pyramidal neuron CA3 and the
MSO neuron. The CA3 is anatomical acronym of Cornu Ammonis, 3rd area
of hippocampus and the MSO, medial superior olive is a nucleus in the
binaural auditory pathway. These model neurons are embodied into three
implementations differing by their levels of complexity, as follows.
1. The first level (macroscopic) unit is phenomenological model with
black-box components containing elementary arithmetic units connected
together, generating spikes as uniform, unitary events [3].
2. The second level (mesoscopic) unit is medium complexity model with
delays and voltages described in experiments [4].
3. The third level (microscopic) unit is biophysically realistic detailed
model based on the anatomical reconstruction of single neuron [1].
In the three levels we describe necessary time step duration and
computational complexity, which has as its practical consequence ration of
simulated versus real time on a prototypical contemporary personal
computer [5,6]. The implementation details may vary. We use GCC, gcc.
gnu.org; www.scholarpedia.org/article/GENESIS[1]; MATLAB (TM), www.
mathworks.com; octave, wiki.octave.org[3]; and other software and
libraries. In order to abstract the implementation details, we theoretically
estimate and recommend the useful range of time step sizes, number of
compartments and computational complexity. This enables us to obtain
asymptotic and converging figures of information transfer rates and
therefore a measure of computational complexity, obtained by the units of
three different levels. Finally, we analyze several models in computational
neuroscience literature (including our previous works). Based on this
analysis, we discuss whether the computational resources were used
efficiently.
Acknowledgements: This work is funded by the Institutional Support for
Long-term Development of Research Organizations (PRVOUK) P 24 to
P. M. at the Charles University in Prague.
References
1. Bower JM, Beeman D: The Book of Genesis New York: Springer Verlag, 2nd
1998.
2. Koch C: (Book Chapter:) Voltage-Dependent Events in the Dendritic Tree.
Biophysics of Computation. Information Processing in Single Neurons Oxford
University Press, New York 1999, 428-452.
3. Bures Z: The stochastic properties of input spike trains control neuronal
arithmetic. Biol Cybern 2012, 106(2):111-122.
4. Bures Z, Marsalek P: On the precision of neural computation with inter-
aural level differences in the lateral superior olive. Brain Res 2013,
1536:16-26.
5. Kuriscak E, Trojan S, Wunsch Z: Model of spike propagation reliability
along the myelinated axon corrupted by axonal intrinsic noise sources.
Physiol Res 2002, 51:205-215.
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 110 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

6. Kuriscak E, Marsalek P, Stroffek J, Wunsch Z: The effect of neural noise on
spike time precision in a detailed CA3 neuron model. Comput Math
Methods Med 2012, 2012:595398, (pp. 1-16).
P159
A neural model of the optomotor system accounts for ordered
responses to decreasing stimulus spatial frequencies
Alex Cope1*, Chelsea Sabo1, Eleni Vasilaki1, Kevin Gurney2, James AR Marshall1
1 Acknowledgements: This work was supported by the Engineering and
Department of Computer Science, University of Sheffield, Sheffield, S10 2TN,
UK; 2Department of Psychology, University of Sheffield, Sheffield, S10 2TN, UK
E-mail: a.cope@sheffield.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P159
In insects the optomotor response produces a motor action to compensate
for unintended body rotation. The response is generally modeled as a
Reichardt-Hassenstein (HSD) or Barlow-Levick (BL) correlation detector, as
anatomical and physiological studies in Drosophila melanogaster have
demonstrated consistent neural pathways and responses in the insect
brain [1]. Recordings from the descending neurons carrying the optomotor
response signal in honeybees indicate an ordering effect for different
stimulus spatial frequencies, with a greater response with decreasing
frequency [2] (see Figure 1A), which is not accounted for by HSD or BL
correlation detectors.
We present a model in the SpineML format of the optomotor system, using
Izhikevich point neurons tuned to match the respective physiological
responses, which is shown in Figure 1B. To examine if the model
reproduces the ordering effect found in the honeybee we performed
simulated annealing on four conductance values in the model, as shown in
Figure 1. The objective function is designed to maximize: correct ordering;
a 10Hz maximum response; and contrast between responses to forward
and reverse motion. A. The data was imported into a commercial software
package (MATLAB 7.14, The MathWorks Inc., Natick, MA, 2012) for analysis
and interpolated onto a 414 grid. A 3D slice of this 4D grid can be seen in
Figure 1C. Spatial frequencies of 32.7, 18.9 and 9.5 Hz are used.
A stable region in which a high value of the objective function, and thus
correct spatial frequency ordering, could be obtained was found. In the
stable region the onset pathway activity is low, leading to offset activity
dominating. A RHD using the model up to the Medulla does not show
correct ordering.
Physical Sciences Research Council [grant number EP/J019534/1].
References
1. Behnia R, Clark DA, Carter AG, Clandinin TR, Desplan C: Processing
properties of ON and OFF pathways for Drosophila motion detection.
Nature 2014, 512:427-430.
2. Ibbotson MR: Evidence for velocity-tuned motion-sensitive descending
neurons in the honeybee. Proc. Biol Sci 2001, 268(1482):2195-2201.
P160
Orientation selectivity in a model of primary visual cortex with and
without orientation map
Soledad Gonzalo Cogno*, Germn Mato
Statistical and Interdisciplinary Physics Group, Instituto Balseiro and Centro
Atmico Bariloche, Bariloche, 8400, Argentina
E-mail: s.gonzalocogno@gmail.com
BMC Neuroscience 2015, 16(Suppl 1):P160
Since its discovery by Hubel and Wiesel in 1959, orientation selectivity
has been observed in every mammal for which the neuronal response
selectivity of primary visual cortex (V1) has been examined. In some

Figure 1(abstract P159) A. Cartoon of ordering effect indicated by honeybee descending neuron responses. Spatial frequency decreases from
blue to green. B. Model diagram showing annealed synapses (coloured, same colours indicate same synaptic conductance). C. Slice of annealing results
(Medulla-Lobula conductance of 0.04nS) showing stable region (green / blue area). D. Model response for a high objective function value (left), and the
corresponding response for an RHD using the model output up to the Medulla
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
animals, like cats and monkeys, anatomically close V1 neurons have
similar preferred orientations, giving rise to maps of orientation
preferences. However, sharp selectivity is also observed in mice, squirrels
and rats, whose V1 has no orientation map. This means that neurons
with different preferred orientations are intermixed. This second scenario
is called salt-and-pepper organization and leads to question the structural
organization of the intracortical connections.
With this study, we intend to analyze the differences between both
scenarios focusing on selectivity properties, and to clarify the effect of
plasticity on orientation selectivity. We study a computational model of
layer 2/3 and a reduced one-dimensional model of orientation selective
neurons, both in the balanced state. We also analyze a plasticity
mechanism that involves spike-timing dependent plasticity (STDP) [1].
This rule implies that if the postsynaptic spike comes after the
presynaptic spike, the connection becomes stronger. For the reverse
order it becomes weaker. Inhibitory synapses are taken as non-plastic.
The selectivity is quantified using the Orientation Selectivity Index (OSI) of
the activity profile.
According to the theory of balanced networks [2] applied to spatially
structured networks [3], we find that for a given profile of the input,
selectivity of the cortical activity is determined by the ratio between the
first Fourier component of the connectivity and its mean value. As a
consequence, for large connectivity salt-and-pepper structures are more
selective than systems with orientation maps: our simulations indicate that
the mean OSI for salt-and-pepper is 0.57, while it is 0.27 for orientation
map. Moreover, for systems with orientation maps selectivity could be
increased by taking connection probabilities that are broader.
Regarding plasticity, we find that under certain conditions STDP can
indeed improve selectivity but it works in a somehow unexpected way,
by decreasing the modulated part of the intracortical connectivity with
respect to the non-modulated part. We find this conclusion to be valid
both for systems with salt-and-pepper organization and with orientation
maps. This can be understood in terms of the relative change between
the background connectivity and the functionally modulated part.
References
1. Bi G, Poo M: Synaptic modfications in cultured hippocampal neurons:
dependence on spike timing, synaptic strength, and postsynaptic cell
type. J Neurosci 1998, 18:10464-10472.
2. van Vreeswijk C, Sompolinsky H: Chaotic balanced state in a model of
cortical circuits. Neural Computation 1998, 10:1321-1372.
3. Rosenbaum R, Doiron B: Balanced networks of spiking neurons with
spatially dependent recurrent connections. Phys Rev X 2014, 4:021039.
P161
When function mirrors structure: how slow waves are shaped by
cortical layers
Cristiano Capone1,2*, Beatriz Rebollo3, Alberto Muoz-Cespdes4,
Paolo Del Giudice2, Maria Victoria Sanchez-Vives3,5, Maurizio Mattia2
1
PhD Program in Physics, Sapienza University, Rome, Italy; 2Istituto Superiore
di Sanit, Rome, Italy; 3IDIBAPS, Barcelona, Spain; 4Universidad Complutense
de Madrid, Madrid, Spain.; 5ICREA, Barcelona, Spain
E-mail: cristiano.capone@infn.iss.it
BMC Neuroscience 2015, 16(Suppl 1):P161
Neuronal spontaneous activity can provide valuable information on the
functional structure of the underlying neuronal network. We focused on
the spontaneous slow-wave activity that is generated and propagated in
cortical slices, aiming at relating their spatio-temporal organization with
the laminar structure of the tissue. We extracted multi-unit activities from
simultaneous recordings obtained by means of an array of 16 electrodes
covering the slice. Although each site displayed alternation between high-
firing (Up) and almost quiescent (Down) states, slow oscillations (SO) were
not spatially homogeneous, showing layer-dependent state durations and
maximum firing rates. A distribution of time lags between electrodes in
the onset and offset of Up states reflected a propagating activity wave. We
found different propagation modes in terms of velocity, direction,
wavefront shape (see Figure 1A), spatial extension and site of origin.
Despite such variability, at the level of single waves, we consistently found
that the head of the wavefront had a spatial distribution forming a strip
almost parallel to the cortical surface (see Figure 1B black line), at a depth
compatible with layer 4 and 5 (L4 and L5). Moreover, this strip widely
Page 111 of 200
overlapped with the region where Up states had the maximum duration
and magnitude (see Figure 1B respectively green and blue line), a region
possibly corresponding to the most excitable part of the network. To test
such hypothesis, we tried to reproduce the properties of this strip using a
large scale simulation of a slice model, with oscillating cortical modules of
spiking neurons arranged in a 2D lattice with nearest neighbor
connectivity. In this homogeneous medium we implemented a non-
monotonic gradient in the connectivity level (see Figure 1C), aiming to
embody the physiological differences in layer excitability, and setting an
overexcited strip (dark circles). We found that the connectivity level
increase was not trivially related to an increase of the speed of wave
propagation, rather this parameter has to be chosen in an optimal range
(see Figure 1C-D) in order to reproduce propagating patterns compatible
with the experiments. Optimal networks were those having a metastable
Down state in which finite-size fluctuations were capable to prime Down-
to-Up transitions. In addition to this, we further predicted that SO features
like maximum firing rate and longest Up state durations should be found
at the same excitable strip.
Conclusions: As the model predicted, we found that strips of early wave
propagation reliably overlapped with the regions where maximum Up state
duration and firing activity occurred, strengthening the duality between
spontaneous activity and network structure. Finally, we matched the
excitable strips with the slice cortical layers as identified by histology,
finding a reliable overlap between such strips and L4 and L5 (see Figure 1E).
Acknowledgements: Supported by EU CORTICONIC contract 600806 and
MINECO BFU2011- 27094.
P162
A novel method for approximating equilibrium single-channel Ca2+
domains
Victor Matveev
New Jersey Institute of Technology, NJ 07030, USA
E-mail: matveev@njit.edu
BMC Neuroscience 2015, 16(Suppl 1):P162
Localized calcium (Ca2+) signals control some of the most fundamental
physiological processes, including synaptic transmission as well as its
activity-dependent plasticity. Computational and mathematical modeling
played a crucial role in the understanding of spatio-temporal Ca 2+
dynamics that drives these processes, and showed that Ca 2+
concentration around a single Ca2+ channel reaches a quasi-stationary
distribution (known as the Ca 2+ nanodomain) within tens of
microseconds after the opening of the channel, and collapses as rapidly
after the closing of the channel. Such localization of Ca 2+ in time and
space is achieved by its rapid diffusion as well as its binding to its
multiple interaction partners collectively called Ca 2+ buffers and Ca 2+
sensors. One of the successes of mathematical modeling was the
development of several analytic approximations describing the
equilibrium concentration of Ca2+ as a function of distance from the
open Ca2+ channel, such as the Rapid Buffering Approximation (RBA), the
Linear Approximation (LA) and the Excess Buffering Approximation (EBA)
[1-4]. Each of these approximations has a particular applicability
parameter regime created by the interplay between the properties of
Ca2+ buffers, in particular their mobility and Ca2+ binding rates, and the
strength of the Ca 2+ current. Here we present a novel approximation
method which does not rely on a specific range of the relevant Ca2+ and
buffer parameters, and is based on matching the low-distance and large-
distance asymptotic behavior of the concentration function. Even at low
orders, the resulting approximation is as accurate as the second-order
RBA and EBA approximations [4], but its validity extends far beyond the
parameter range of applicability of RBA and EBA. The usefulness of the
resulting approximation is two-fold: first, together with the previously
developed approximations, the novel method could provide a deeper
intuition into the dependence of Ca2+ nanodomain properties on the
relevant buffering parameters, and second, it constitutes an efficient
numerical approximation tool in the modeling of the Ca 2+ signals
underlying presynaptic and postsynaptic phenomena.
References
1. Neher E: Usefulness and limitations of linear approximations to the
understanding of Ca2+ signals. Cell Calcium 1998, 24(5-6):345-357.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P161) A. Wavefronts for 2 modes of propagation. B. Average strips where wavefronts propagate earlier (black), and where Up states
have maximum duration (green) and magnitude (blue). C. Modulation of the connectivity parameter in the model. D. Nullclines under mean-field
approximation varying levels of connectivity. and C are average firing rate and fatigue level, respectively. Circles, fixed points. Dark to light gray, different
excitability levels as in C, respectively. E. Example match between strip of early wave propagation and slices layers
2. Smith GD, Wagner J, Keizer J: Validity of the rapid buffering
approximation near a point source of calcium ions. Biophys J 1996,
70(6):2527-2539.
3. Bertram R, Smith GD, Sherman A: Modeling study of the effects of
overlapping Ca2+ microdomains on neurotransmitter release. Biophys J
1999, 76(2):735-750.
4. Smith GD, Dai LX, Miura RM, Sherman A: Asymptotic analysis of buffered
calcium diffusion near a point source. Siam J Appl Math 2001, 61(5):1816-1838.
P163
How central inputs and force and velocity feedbacks determine
motoneurons activity during voluntary hand movements
Alberto Mazzoni1*, Francesco M Petrini2,3, Jacopo Rigosa1,2,3,
Marco Capogrosso2,3, Stanisa Raspopovic1,2,3, Silvestro Micera1,2,3
1
The BioRobotics Institute, Scuola Superiore SantAnna, Pisa, 56026, Italy;
2
Bertarelli Foundation Chair in Translational NeuroEngineering, Institute of
Bioengineering, School of Engineering, Ecole Polytechnique Federale de
Lausanne, Lausanne, Switzerland; 3Center for Neuroprosthetics, Ecole
Polytechnique Federale de Lausanne, Lausanne, Switzerland
E-mail: a.mazzoni@sssup.it
BMC Neuroscience 2015, 16(Suppl 1):P163
Human hand motion is the result of a complex interplay of motoneurons
dynamics, central inputs and feedbacks from the muscles activity. A
complete picture of this interplay is still missing, also due to the difficulty
of recording motoneuron activity. Thanks to a novel recording method we
were able to observe motoneurons spiking activity in the human median
nerve during voluntary hand movements. We characterized then the
neural dynamics associated to force-varying tasks and to fixed velocity
tasks involving different muscles. We used these results to develop a
spiking neuron model for the interpretation of the observed relationship
between motoneurons firing rate and muscle activity features shedding
light on the neural interactions underlying control of hand movements.
The model builds on previous studies [1,2] i) to define a common drive to
motoneurons proportional to the target force and ii) to define a
Page 112 of 200
relationship between the motoneurons firing rate and the resulting
muscular force. Starting from these works we introduced a number of
modifications (Figure 1). First, we modeled single neurons as Regular
Spiking Izhikevich neurons [3] rather than impulse generators, in order to
take into account firing rate adaptation and to monitor relevant
biological parameters. Second, we considered an unsupervised and
biologically sound force feedback: instead of being determined by the
difference between the prescribed target force and the actual force
generated by the muscles, the feedback is determined exclusively by the
resulting force. This feedback emulates the input to motoneurons sent by
Ib afferent fibers. Finally, we included in the model also a specific
feedback associated to the kinematic of the movements [4], analogously
to the feedback contribution from Ia and II afferent fibers. We found that
experimental results were reproduced only when i) central input was
completely determined by the requested force; ii) force and kinematic
feedback were respectively inhibitory and excitatory. Interestingly, we
also found that a weak adaptation can account for a large fraction of the
experimentally observed firing rate saturation at high forces even in the
absence of feedback, while feedbacks are needed for the fine modulation
of the outputs. Model predictions on the central input and the feedback
dynamics will be tested in future experiments isolating the different
components of the reflex network.
Acknowledgements: This work was supported by FP7-602547 EPIONE,
FP7-611687 NEBIAS, Italian Ministry of Health grant NEMESIS, Italian
Ministry of University and Research grant HANDBOT, and by Scuola
Superiore SantAnna.
References
1. Fuglevand AJ, Winter DA, Patla AE: Models of recruitment and rate coding
organization in motor-unit pools. J Neurophysiol 1993, 70:2470-2488.
2. Contessa P, De Luca CJ: Neural control of muscle force: indications from
a simulation model. J Neurophysiol 2013, 109:1548-1570.
3. Izhikevich EM: Simple model of spiking neurons. IEEE Transactions on
neural networks 2003, 14:1569-1572.
4. Maltenfort MG, Burke RE: Spindle model responsive to mixed fusimotor
inputs and testable predictions of feedback effects. IEEE Transactions on
neural networks 2003, 14:1569-1572.
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 113 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

Figure 1(abstract P163) Illustrative scheme of network structure

of cortical column models [2], knowledge of connectivity [3,4], scalable

P164
Induction of long-term potentiation and depression in individual
synapses of CA1 pyramidal neurons
Rosanna Migliore*, Giada De Simone, Michele Migliore
Institute of Biophysics, National Research Council, via Ugo La Malfa 153,
90146 Palermo, Italy
E-mail: rosanna.migliore@cnr.it
BMC Neuroscience 2015, 16(Suppl 1):P164
Long-Term Potentiation (LTP) and Depression (LTD), the two major forms of
long-lasting synaptic plasticity in the mammalian neurons, represent a basic
step to understand neuronal development, circuit reorganization, learning
and memory mechanisms. Experimental studies on LTP and LTD are usually
performed using in vitro or in vivo preparations, exploiting repetitive and
properly patterned stimulation protocols to induce long-lasting changes in
the strength of a synaptic connection. Although the underlying molecular
mechanisms are starting to be unraveled and are under intense experimental
and theoretical scrutiny, especially in the CA1 region of the hippocampus,
experimenters often report problems in using standard induction protocols to
obtain consistent results, especially for LTD in vivo. We hypothesize that a
possible source of confusion in interpreting the results, from any given
experiment on synaptic plasticity, can be the intrinsic limitation of the
experimental techniques, which cannot take into account the actual state and
peak conductance of the synapses before the conditioning protocol.
Within this context, using biophysical models of synaptic plasticity and
hippocampal CA1 pyramidal neurons, we investigate the relation between
what is observed at the soma and LTP/LTD induction patterns, pre-
conditioning synaptic strength, and dendritic location.
Our model and the results pointed out that the outcome of an
experiment, testing the amount of synaptic LTP/LTD plasticity that can be
induced, strongly depends at least on the initial synaptic state and peak
conductance. In addition, the model explains why LTD induction may be
more critical to be obtained, with respect to LTP, especially in vivo, and
suggests experimentally testable predictions on the stimulation protocols
that may be more effective.
Acknowledgements: We thank the Cineca consortium (Bologna, Italy) for
granting access to their IBM BlueGene/Q FERMI system.
The research leading to these results has received funding from the
European Union Seventh Framework Program (grant No. FP7/2007-2013)
under grant No. 604102 (Human Brain Project).
simulation tools [NEST] and robust population statistic techniques [5].
Vasoactive intestinal peptide-positive (VIP+) interneurons are one of the
major inhibitory cell types in the cortex [1], and their excitability has been
found to be associated with various endogenous factors such as brain
states and top-down signals. Zhang et al. [6] recently found that top-down
signals originated in the cingulate (Cg) strongly innervate VIP +
interneurons of V1, indicating the critical role of VIP+ cells in attentional
gain modulation. We studied their effect on top-down signals using both
analytical and computational models. Using a coarse-grained firing rate
model for the cell types in the superficial layers we found that the
reported cell-type specific connectivity allows top-down inputs to VIP+
interneurons to effectively disinhibit pyramidal cell activity. Due to the
limitations of the firing rate model, we utilized large scale computational
models to investigate mechanisms underlying attentional gain modulation
in an entire network. Specifically, we adopted the multiple column model
proposed by Wagatsuma et al. [4] and modified them by incorporating
three inhibitory cell types into the superficial layers of the model. In our
model we considered parvalbumin-positive (PV+), SST+ and VIP+ inhibitory
cells. We also constructed the data-driven cell-type specific intercolumnar
connections between columns. Our model was capable of reproducing
multiplicative gain modulation (see Figure 1), and our simulation results
suggest that non-specific activation of VIP cells is sufficient for generating
multiplicative gain modulation.
We also note that PV+ cell-mediating inhibition across columns is critical
for regulating the responses over small spatial scales: e.g. induced by
non-preferred stimulus (orientation). SST+ cell mediate inhibition across
columns and allow contextual visual processing in V1.

P165
Cell-type specific connectivity accounts for diverse in vivo functional
roles of inhibitory neurons in V1
Jung H Lee, Stefan Mihalas*
Allen Institute for Brain Science, Seattle, WA 98103, USA
E-mail: stefanm@alleninstitute.org
BMC Neuroscience 2015, 16(Suppl 1):P165

Inhibitory neurons have a large diversity [1], however functional roles of

diverse subtypes have not been elucidated. We aim to help systematically Figure 1(abstract P165) Simulated tuning curve for pyramidal
addressing this question using large scale modeling and coarse graining neurons as a function of top-down inputs to VIP cells. These inputs
between scales. Mesoscopic models of in vivo activity starting from allow a multiplicative gain modulation of the bottom-up input
structure have only recently become feasible making use of development
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
References
1. Rudy B, Fishell G, Lee S, Hjerling-Leffler J: Three groups of interneurons
account for nearly 100% of neocortical GABAergic neurons.
Developmental Neurobiology 2011, 71(1):45-61, doi:10.1002/dneu.20853.
2. Wagatsuma N, Potjans TC, Diesmann M, Sakai K, Fukai T: Spatial and
Feature-Based Attention in a Layered Cortical Microcircuit Model. PLoS
One 2013, 8(12):e80788, doi:10.1371/journal.pone.0080788.
3. Pfeffer CK, Xue M, He M, Huang ZJ, Scanziani M: Inhibition of inhibition in
visual cortex: the logic of connections between molecularly distinct
interneurons. Nature Neuroscience 2013, 16(8):1068-1076, doi:10.1038/nn.3446.
4. Oh SW, et al: A mesoscale connectome of the mouse brain. Nature 2014,
508(7495):207-14, Apr 10; doi: 10.1038/nature13186. Epub 2014 Apr 2.
5. Iyer R, Menon V, Buice M, Koch C, Mihalas S: The influence of synaptic
weight distribution on neuronal population dynamics. PLoS Comput Biol
2013, 9(10):e1003248, Oct; doi: 10.1371/journal.pcbi.1003248.
6. Zhang S, Xu M, Kamigaki T, Hoang Do JP, Chang WC, Jenvay S,
Miyamichi K, Luo L, Dan Y: Long-range and local circuits for top-down
modulation of visual cortex processing. Science 345(6197):660-665,
doi:10.1126/science.1254126.
P166
Reconstructing the directionality of coupling between cortical
populations with negative phase lag
Fernanda S Matias1*, Leonardo L Gollo2, Pedro V Carelli3, Mauro Copelli3,
Claudio R Mirasso4
1
Instituto de Fsica, Universidade Federal de Alagoas, Macei, AL 57072-900,
Brazil; 2Systems Neuroscience Group, Queensland Institute of Medical Research,
Brisbane QLD 4006, Australia; 3Departamento de Fsica, Universidade Federal de
Pernambuco, Recife PE 50670-901, Brazil; 4Instituto de Fisica Interdisciplinar y
Sistemas Complejos, CSIC-UIB, Campus Universitat de les Illes Balears, E-07122
Palma de Mallorca, Spain
E-mail: fernanda@fis.ufal.br
BMC Neuroscience 2015, 16(Suppl 1):P166
Understanding how information is processed in the brain is one of the key
areas in neuroscience research.Different tools have been employed to
reconstruct directional influence and to infer the effective connectivity
between distinct brain regions. Particularly, it has been shown [1] that in
non-linear delay-coupled oscillating systems exhibiting a negative phase
lag, Granger causality (GC) might not provide the correct direction of
information flow (from the driver to the receiver). Such systems have been
studied before in the theoretical framework of Anticipated Synchronization
(AS) developed in the field of dynamical systems [2]. This counterintuitive
synchronization regime can be a stable solution of two dynamical systems
coupled in a master-slave (driver-receiver) configuration when the slave
receives a negative delayed self-feedback. Recently, it has been shown that
unidirectional coupled neuronal population models can also exhibit AS [3].
In these cortical like populations the delayed feedback has been replaced
by a dynamical inhibitory loop mediated by interneurons. Here we show
that in these biologically plausible models, GC provides the correct
directionality of the coupling for both positive and negative phase
differences. In fact, when compared to experimental data in the primate
cortex our model reproduces experimental phase lags, coherence spectra
and GC spectra [3].
References
1. Vakorin VA, Krakovska O, McIntosh AR: On Complexity and Phase Effects
in Reconstructing the Directionality of Coupling in Non-linear Systems.
Directed Information Measures in Neuroscience Springer Berlin Heidelberg
2014, 137-158.
2. Voss HU: Anticipating chaotic synchronization. Phys Rev E 2000, 61:5115-5119.
3. Matias FS, Gollo LL, Carelli PV, Bressler SL, Copelli M, Mirasso CR: Modeling
positive Granger causality and negative phase lag between cortical
areas. NeuroImage 2014, 99:411-418.
P167
On the basic mechanisms of anticipated synchronization in neuronal circuits
Fernanda Matias1*, Ana Paula Millan2, Luis Martinez3, Santiago Canals3,
Pedro Carelli4, Mauro Copelli4, Claudio R Mirasso5
1
Instituto de Fsica, Universidade Federal de Alagoas, 57072-900 Macei,
Brazil; 2Departamento de Electromagnetismo y Fsica de la Materia,
Page 114 of 200
Universidad de Granada, 18071 Granada, Spain; 3Instituto de Nuerociencias
de Alicante, 03690 Sant Joan dAlacant, Spain; 4Departamento de Fsica,
Universidade Federal de Pernambuco, Recife PE 50670-901, Brazil; 5Instituto
de Fisica Interdisciplinar y Sistemas Complejos, CSIC-UIB, Campus Universitat
de les Illes Balears, E-07122 Palma de Mallorca, Spain
E-mail: claudio@ifisc.uib-csic.es
BMC Neuroscience 2015, 16(Suppl 1):P167
Anticipated synchronization (AS) is an anti-intuitive phenomenon that can
occur in two coupled dynamical systems when there is a dominant
connectivity between the elements. AS occurs when a dynamical system
A dominantly connects to another system B and B synchronously pulses
before A does. It has been recently shown [1,2] that AS can occur in a
model of coupled Hodgkin-Huxley (HH) neurons and even in neuron
populations. Recently, this astonishing regime has been observed in some
cortical circuits of monkeys when performing a visual discrimination task
[2]. However, the basic mechanisms for this synchronization to occur are
still unclear. In this communication we analyze a circuit of excitatory and
inhibitory HH neurons as well as neurons populations and find, analyzing
individual responses as well as phase response curves, that inhibitory
neurons can control the transition between delayed and anticipated
synchronization.
References
1. Matias FS, Carelli PV, Mirasso CR, Copelli M: Anticipated synchronization in
a biologically plausible model of neuronal motifs,. Phys Rev E 2011,
84:021922.
2. Matias FS, Gollo LL, Carelli P, Bressler S, Copelli M, Mirasso CR: Modelling
positive Granger Causality and negative phase lag between cortical
areas. Neuroimage 2014, 99:411-418.
P168
Information transfer by local field potentials in the hippocampal
formation
Maria Constantinou*, Daniel Squirrell, John Gigg, Marcelo A Montemurro
Faculty of Life Sciences, University of Manchester, Manchester, M13 9PT, UK
E-mail: maria.constantinou@manchester.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P168
Extracellular electrical potential oscillations recorded as local field
potentials (LFPs) in the hippocampal formation are thought to be involved
in cognitive processes such as working memory retention, memory
consolidation and spatial navigation. Recent studies have shown that
combining spikes with LFP phase can increase the information content of
spikes [1,2] and thus suggest LFP oscillations have the capacity to convey
information. LFP oscillations within specific frequency bands can interact,
for example by phase-phase and phase-amplitude coupling. We
hypothesise that these LFP interactions can transfer information between
neural networks. To test this hypothesis, we analyse multi-channel
recordings of simultaneous LFPs from hippocampal area CA1 and the
subiculum of urethane-anaesthetised rodents. Anatomical connections
between these two regions in the hippocampal formation predict
information in this system flows in nested loops along separate
projections [3]. We use advanced neurocomputational methods to
determine how information flows within the CA1-subicular circuit by
interactions of LFP rhythms. The results of correlation and coherence
analyses of LFPs recorded from multiple sites suggest that LFP rhythms
can transmit information within and between area CA1 and the subiculum.
However, these methods alone are not enough to determine the
oscillatory activity in which region drives activity in other regions, that is in
which direction information flows. We use transfer entropy [4], which is an
information theoretic method that can capture directionality, to quantify
information transfer between area CA1 and subiculum. We show that
information flow within the CA1-subicular circuit is bi-directional and
follows a pattern of feedforward and feedback loops. Our results suggest
that LFP interactions can route information at millisecond timescales along
the anatomical connections in the hippocampal formation to achieve
cognitive processing.
References
1. Montemurro MA, Rasch MJ, Murayama Y, Logothetis NK, Panzeri S: Phase-
of-firing coding of natural visual stimuli in primary visual cortex. Curr Biol
2008, 18(5):375-380.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
2. Kayser C, Montemurro MA, Logothetis NK, Panzeri S: Spike-phase coding
boosts and stabilizes information carried by spatial and temporal spike
patterns. Neuron 2009, 61(4):597-608.
3. Gigg J: Constraints on hippocampal processing imposed by the
connectivity between CA1, subiculum and subicular targets. Behav Brain
Res 2006, 174(2):265-271.
4. Schreiber T: Measuring information transfer. Phys Rev Lett 2000,
85(2):461-464.
P169
ROS-MUSIC toolchain for spiking neural network simulations in a
robotic environment
Philipp Weidel1*, Renato Duarte1, Karolna Korvasov1, Jenia Jitsev1,
Abigail Morrison1,2,3
1
Institute of Advanced Simulation (IAS-6) & Institute of Neuroscience and
Medicine (INM-6), Forschungszentrum Juelich, 52425 Juelich, Germany;
2
Faculty of Psychology, Institute of Cognitive Neuroscience, Ruhr-University
Bochum, 44801 Bochum, Germany; 3Simulation Laboratory Neuroscience -
Bernstein Facility for Simulation and Database Technology, Institute for
Advanced Simulation, Jlich Aachen Research Alliance, Jlich Research
Center, Jlich, Germany
E-mail: p.weidel@fz-juelich.de
BMC Neuroscience 2015, 16(Suppl 1):P169
Studying a functional, biologically plausible neural network that performs
a particular task is highly relevant for progress in both neuroscience and
machine learning. Most tasks used to test the function of a simulated
neural network are still very artificial and thus too narrow, providing only
little insight into the true value of a particular neural network architecture
under study. For example, many models of reinforcement learning in the
brain rely on a discrete set of environmental states and actions [1]. In
order to move closer towards more realistic models, modeling studies
have to be conducted in more realistic environments that provide
complex sensory input about the states. A way to achieve this is to
provide an interface between a robotic and a neural network simulation,
such that a neural network controller gains access to a realistic agent
which is acting in a complex environment that can be flexibly designed
by the experimentalist.
To create such an interface, we present a toolchain, consisting of already
existing and robust tools, which forms the missing link between robotic
and neuroscience with the goal of connecting robotic simulators with
neural simulators. This toolchain is a generic solution and is able to
combine various robotic simulators with various neural simulators by
connecting the Robot Operating System (ROS) [2] with the Multi-
Simulation Coordinator (MUSIC) [3]. ROS is the most widely used
middleware in the robotic community with interfaces for robotic
simulators like Gazebo, Morse, Webots, etc, and additionally allows the
users to specify their own robot and sensors in great detail with the
Unified Robot Description Language (URDF). MUSIC is a communicator
between the major, state-of-the-art neural simulators: NEST, Moose and
NEURON. By implementing an interface between ROS and MUSIC, our
toolchain is combining two powerful middlewares, and is therefore a
multi-purpose generic solution.
One main purpose is the translation from continuous sensory data,
obtained from the sensors of a virtual robot, to spiking data which is
passed to a neural simulator of choice. The translation from continuous
data to spiking data is performed using the Neural Engineering Framework
(NEF) proposed by Eliasmith & Anderson [4]. By sending motor commands
from the neural simulator back to the robotic simulator, the interface is
forming a closed loop between the virtual robot and its spiking neural
network controller.
To demonstrate the functionality of the toolchain and the interplay
between all its different components, we implemented one of the
vehicles described by Braitenberg [5] using the robotic simulator Gazebo
and the neural simulator NEST.
In future work, we aim to create a testbench, consisting of various
environments for reinforcement learning algorithms, to provide a
validation tool for the functionality of biological motivated models of
learning.
Page 115 of 200
References
1. Jenia Jitsev, Morrison Abigail, Tittgemeyer Marc: Learning from positive
and negative rewards in a spiking neural network model of basal
ganglia. Neural Networks (IJCNN), The 2012 International Joint Conference on.
IEEE 2012.
2. Morgan Quigley, et al: ROS: an open-source Robot Operating System..
ICRA workshop on open source software 2009, 3(3.2).
3. Mikael Djurfeldt, et al: Run-time interoperability between neuronal
network simulators based on the MUSIC framework. Neuroinformatics
2010, 8.1:43-60.
4. Chris Eliasmith, Charles H. Anderson: Neural engineering: Computation,
representation, and dynamics in neurobiological systems. MIT press 2004.
5. Valentino Braitenberg: Vehicles: Experiments in synthetic psychology. MIT
press 1986.
P170
SPIKE-Synchronization: a parameter-free and time-resolved coincidence
detector with an intuitive multivariate extension
Thomas Kreuz*, Nebojsa Bozanic, Mario Mulansky
Institute for Complex Systems, CNR, Sesto Fiorentino, Italy
E-mail: thomas.kreuz@cnr.it
BMC Neuroscience 2015, 16(Suppl 1):P170
Techniques for recording large-scale neuronal spiking activity are
developing very fast. This leads to an increasing demand for algorithms
capable of analyzing large amounts of experimental spike train data. One
of the most crucial and demanding tasks is the identification of similarity
patterns with high temporal resolution and across different spatial scales.
To address this task, in recent years three time-resolved measures of
spike train synchrony have been proposed, event synchronization [1], the
ISI-distance [2], and the SPIKE-distance [3].
Here we present SPIKE-synchronization [4], an improved and simplified
extension of event synchronization with a more intuitive interpretation
which holds for both the bivariate and the multivariate case. SPIKE-
synchronization quantifies the degree of synchrony from the relative
number of quasi-simultaneous appearances of spikes. Since it builds on
the same bivariate and adaptive coincidence detection that was used for
event synchronization, SPIKE-synchronization is parameter- and scale-free
as well. This makes it easy to handle and allows for an objective estimation
of neuronal synchronization. In contrast to the ISI- and the SPIKE-distance,
SPIKE-synchronization is a measure of similarity. It is zero if and only if the
spike trains do not contain any coincidences, and reaches one if and only
if each spike in every spike train has one matching spike in all the other
spike trains.
We investigate the properties of SPIKE-synchronization and compare it
against other time-resolved measures such as the Peri-Stimulus Time
Histogram (PSTH) and the ISI- and the SPIKE-distance [4,5]. We use
simulated data to verify its usefulness and explore its performance on
real data.
Together with the ISI-distance and the SPIKE-distance, SPIKE-
Synchronization is implemented in both the Matlab-based graphical user
interface SPIKY and the Python library PySpike [6]. Both packages provide
ample documentation as well as platforms for user feedback. SPIKY even
comes with an interactive Facebook page (https://www.facebook.com/
SPIKYgui) and a YouTube channel (https://www.youtube.com/user/
SPIKYgui1) which includes movies demonstrating both the measures and
the GUI.
Acknowledgements: We acknowledge funding support from the
European Commission through Marie Curie Initial Training Network
Neural Engineering Transformative Technologies (NETT), project 289146
and through the European Joint Doctorate Complex oscillatory systems:
Modeling and Analysis (COSMOS), project 642563 (TK).
References
1. Quian Quiroga R, Kreuz T, Grassberger P: Event Synchronization: A simple
and fast method to measure synchronicity and time delay patterns. Phys
Rev E 2002, 66:, 041904.
2. Kreuz T, Haas JS, Morelli A, Abarbanel HDI, Politi A: Measuring spike train
synchrony. J Neurosci Methods 2007, 165:, 151-161.
3. Kreuz T, Chicharro D, Houghton C, Andrzejak RG, Mormann F: Monitoring
spike train synchrony. J Neurophysiol 2013, 109:1457.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
4. Kreuz T, Mulansky M, Bozanic N: SPIKY: A graphical user interface for
monitoring spike train. Arxiv 2015, 1410.6910v2 (Submitted to
JNeurophysiol).
5. Mulansky M, Bozanic N, Sburlea A, Kreuz T: A guide to time-resolved and
parameter-free measures of spike train synchrony. Arxiv 2015, 1502.02027
(Submitted to IEEE).
6. , Source codes of SPIKY and PySpike are available at http://www.fi.isc.cnr.it/
users/thomas.kreuz/sourcecode.html and https://github.com/
mariomulansky/PySpike, respectively.
P171
Novel perspective on field recordings in zebrafish models of epilepsy
Adriana Dabacan1,2*, Sorana Ciura3, Edor Kabashi3, Hortense de Calbiac3,
Raul Muresan1
1
Coneural, Romanian Institute of Science and Technology, Cluj Napoca,
Romania; 2Basis of Electronics, UTCN, Cluj-Napoca, Romania; 3Amyotrophic
Lateral Sclerosis: From Genetics to Treatment, ICM, Paris, France
E-mail: dabacan@rist.ro
BMC Neuroscience 2015, 16(Suppl 1):P171
Research in Epilepsy relies strongly on animal models, either for describing
genetic conditions involved in the disease or for testing potential drugs
that might alleviate the symptoms [1].
In this study, we looked at the differential effect of Pentylenetetrazole
(PTZ), a well-known epileptogenic drug [2], on 5-6 dpf zebrafish larvae of
two different genetic conditions: a mutant line, where a gene known to be
involved in focal epilepsy was specifically knocked down, as well as control
mismatch oligonucleotide-injected zebrafish.
Figure 1(abstract P171) Correlation histograms on events from zebrafish field recordings: Top: Autocorrelation of mutant (A) versus control (B)
animals; Bottom: Scaled autocorrelation with scale segment of 1000s of mutant (C) versus control (D) animals
Page 116 of 200
The zebrafish larvae were scored for phenotypic features, including
hyperactivity and aberrant locomotion. These phenotypic features were
present in all mutant fish recorded, but were absent in the mismatch
control. We acquired electrophysiological field recordings from the optic
tectum before and after PTZ application and selected a low frequency
band of the signal (0.05 - 0.5 Hz). Troughs, representing epileptic events,
were extracted and a wide range of analyses were applied: total event
count, event time histogram, event duration distributions, inter-event-
interval (IEI) distribution, classical and scaled autocorrelations on the field
and event signals [3].
Traditional analysis provided a quantitative evaluation of PTZ-induced
epileptic events: an increase in number of epileptic events was observed
in mutant fish with respect to control, and the event time histogram
showed a more abrupt increase at 32-37 min. after PTZ application,
whereas in the control case, the increase was constant throughout the
response period. Compared to controls, in mutant fish, PTZ application
led to a larger increase in the number of short (0.5-1.5 s) events with
small IEI (< 10s).
Correlation analysis revealed qualitative information about the effect of
PTZ on field recordings: The scaled autocorrelation (scale segment of 1s)
on the field signal exhibited oscillatory components around 0.4 Hz in all
conditions, but mutant fish exhibited frequency variability with time after
PTZ application and variability across animals, leading to a low average
oscillation power. In controls the frequency was robustly locked at 0.4 Hz.
Autocorrelation histograms computed on the extracted events evidenced
a wide refractory period (~ 4 s, see Figure 1A), followed by baseline in
mutant fish recordings, whereas control animals exhibited a narrow
refractory period (~ 2 s, see Figure 1B), followed by a secondary peak and
a slow modulation (~ 35 s). Scaled autocorrelation histograms (scale
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
segment of 1000s) showed constant correlation decrease with increased
lag in the control fish, but not in mutant fish, where correlation for larger
lags fluctuated around zero (see Figure 1C and 1D).
The characteristics revealed by correlation analysis suggest potentially
different mechanisms underlying PTZ-induced epileptic events. We
therefore propose that, in addition to the traditional statistics on epileptic
events, looking at the temporal characteristics and correlation structure of
field recordings may lead to a better classification and understanding of
the mechanisms underlying epileptogenesis in various models of epilepsy.
Acknowledgements: This paper is supported by the Sectoral Operational
Programme Human Resources Development POSDRU/159/1.5/S/137516
financed from the European Social Fund and by the Romanian Government.
References
1. Grone B, Baraban S: Animal models of epilepsy: Legacies and new
directions. bioRxiv 2015, 013136.
2. Baraban SC, Taylor MR, Castro PA, Baier H: Pentylenetetrazole induced
changes in zebrafish behavior, neural activity and c-fos expression.
Neuroscience 2005, 131(3):759-768.
3. Nikolic D, Muresan RC, Feng W, Singer W: Scaled correlation analysis: a
better way to compute a cross-correlogram. European Journal of
Neuroscience 2012, 35(5):742-762.
P172
Shaping pathological cortical dynamics with high-frequency
neurostimulation
Jrmie Lefebvre1*, Micah M Murray1,2
1
Laboratory for Investigative Neurophysiology (The LINE), Centre Hospitalier
Universitaire Vaudois, Lausanne, 1011, Switzerland; 2EEG Brain Mapping Core,
Centre for Biomedical Imaging (CIBM), 1011 Lausanne, Switzerland
E-mail: jeremie.lefebvre@hotmail.com
BMC Neuroscience 2015, 16(Suppl 1):P172
Disrupted neural synchrony is a hallmark pattern observed in numerous
pathological states, such as dementia, depression, Parkinsons disease,
epilepsy, and schizophrenia. Patients with brain lesions and conditions
involving brain damage commonly display abnormal slow-wave EEG/MEG
oscillatory activity that further shapes responses at rest and in response
to stimuli [1]. Such pathological dysrhythmia usually takes the form of a
gradual shift between fast cyclic activity in favor of slower frequency
bands and strongly correlates with the degree of neurocognitive
impairment [2]. Despite the diversity in the underlying pathology,
dysrhythmia is a robust, consistent observation across many of these
conditions, suggesting that synchrony disruption might alone be
responsible of some of the symptoms. Can these defective oscillations be
fixed?
In parallel and despite the surge of interest in neurostimulation
techniques (e.g. DBS, TMS, tDCS) to treat brain disorders and/or
manipulate brain activity, little computational insight has been gathered
about the influence of electromagnetic drive on the dynamics of neural
populations. Notably, the interference patterns between stimulation
signals and ongoing oscillatory states are a highly debated yet still poorly
understood problem [3]. How does neurostimulation interact with
synchronous dynamics in the presence of pathological dysrhythmia? Can
one actively tune ongoing oscillatory neural activity in order to treat
neurophysiological disorders and alleviate symptoms of brain-related
dysfunctions?
In this work, we analyze the influence of high-frequency stimulation on
synchronous dynamics in heterogeneous networks. Using a non-linear
and sparsely connected network of cortical neurons with finite
conduction velocity, we examine the functional impact of local damage
on network oscillatory dynamics, using both spiking and mean-field
descriptions, and thus assess the influence of local neural density
depletion on the genesis of pathological dysrhythmia. Our model reveals
that deceleration of rhythmic activity scales with damage size, and
further pairs with a concomitant decrease in spectral power. Then,
building on recent findings about input-mediated tuning of synchronous
oscillations and state of the art non-linear analysis tools [4], we provide
insights about how weak, high-frequency neurostimulation can be used
to accelerate and potentially restore oscillatory activity to healthy levels.
By doing so, we propose a framework in which peculiar neurostimulation
patterns, such as those accessible via TMS or tDCS used in conjunction
Page 117 of 200
with EEG, could bi-directionally regulate the frequency and power
expressed by synchronous cortical networks. These novel findings are
further shown to be distinct from the more common resonance and
entrainment phenomena in which periodic stimuli enhance or even
replace ongoing neural dynamics.
By addressing the implications of structural heterogeneity on synchronous
dynamics and providing novel ways of compensating for them, our work
outlines the key role played by axonal delays on healthy neural activity.
These developments further provide key analytical and numerical insights
about how pulsatile stimulation can be used to shape ongoing cyclic
activity in the healthy and damaged brain.
Financial support for this work has been provided by: National Science and
Engineering Council of Canada, Swiss National Science Foundation.
References
1. van Wijk BCM, Willemse RB, Vandertop WP, Daffertshofer A: Slowing of M1
oscillations in brain tumor patients in resting state and during
movement. Clin Neurophysiol 2012, 123(11):2212-2219.
2. Rossini PM, Rossi S, Babiloni C, Polich J: Clinical neurophysiology of aging
brain: from normal aging to neurodegeneration. Prog Neurobiol 2007,
83(6):375-400.
3. Thut G, Veniero D, Romei V, Miniussi C, Schyns P, Gross J: Rhythmic TMS
causes local entrainment of natural oscillatory signatures. Curr Biol 2011,
21(14):1176-1185.
4. Lefebvre J, Hutt A, Knebel JF, Whittingstall K, Murray MM: Stimulus
Statistics Shape Oscillations in Nonlinear Recurrent Neural Networks.
J Neurosci 2015, 35(7):2895-2903.
P173
Synaptic inputs are tuned to match intrinsic properties to maintain
phase in oscillatory neural networks
Haroon Anwar1,2*, Jordan Storms1,2, Farzan Nadim1,2
1
Federated Department of Biological Sciences, New Jersey Institute of
Technology, Newark, NJ 07102, USA; 2Rutgers University-Newark, Newark, NJ
07102, USA
E-mail: hanwar@njit.edu
BMC Neuroscience 2015, 16(Suppl 1):P173
Rhythmic motor activity requires accurate temporal coordination of
neurons controlling the muscles to produce a stable output pattern. Such
coordinated activity manifests itself as the maintenance of the relative
activity phases of the central pattern generating (CPG) neurons producing
the rhythmic motor activity, in spite of a large range of network
frequencies [1,2]. In many systems, this phase maintenance occurs even
when the behavior is compared across different animals [1,2]. Surprisingly
though, both the synaptic strengths [2] and voltage-gated ionic
conductances [3] vary extensively across animals, leading to the question
of how stability at the network level can arise despite such variability in
the neuronal and synaptic components. We examine the hypothesis that
phase maintenance in oscillatory networks involves a precise matching of
the synaptic inputs to the neurons with the intrinsic properties of the
neuron using the pyloric CPG network in stomatogastic nervous system
(STNS) of crustaceans.
We used the biophysical model description of the pyloric follower LP
bursting neuron [4] and a multi objective genetic algorithm to find
multiple optimal models with varying levels of maximal conductance,
each of which produced the appropriate bursting features (within the
measured biological range) in response to the same periodic synaptic
input. Additionally, all 24 optimal models exhibited tonic firing in the
absence of synaptic input, as observed in biological LP neurons when
synapses are blocked. We then removed the synaptic input to the LP
neuron models and simulated dynamic clamp experiments by injecting
34 different synaptic conductance waveforms (which were recorded
experimentally during ongoing rhythm) into the LP models, which, in
most cases, switched the LP models from tonic firing to bursting. These
simulations were repeated with the waveforms scaled to 500, 750, 1000,
1250, 1500, 1750 and 2000 ms, thus driving the LP neuron at these
periods.
Our results show that when different LP models are injected with a
unique synaptic waveform of fixed maximal amplitude the output varies
greatly. Similarly, when a unique LP model is injected with different
synaptic waveforms of fixed maximal amplitude the output also varies
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 118 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

greatly. This indicates that not only the maximal amplitude of

conductance but the other features of synaptic waveforms also play an
important role in shaping the activity of LP neurons. Consequently, a
distinct synaptic waveform would be required for each of the different LP
model neurons to produce the same output features in LP neurons. Using
this approach, we aim to explore what features of synaptic conductance
are important determinant of changes in output features and how those
synaptic properties match with different ionic conductances to maintain
stereotypical output of pyloric network across animals. We further aim to
confirm these results using dynamic clamp experiments in biological LP
neurons with different synaptic waveforms. Our results also show that
when the cycle period of each synaptic input is varied and applied to
individual LP models, the phase of LP model neuron advances as
expected. We aim to repeat all simulations with different maximal
conductance values of synaptic waveforms. This will reveal the changes
in synaptic strength required for the phase maintenance, when the cycle
period is varied.
Acknowledgements: This project was supported by NIH grant
MH060605.
References
1. Bucher D, Prinz AA, Marder E: Animal-to-Animal Variability in Motor
Pattern Production in Adults and during Growth. J Neurosci 2005,
25(7):1611-1619.
2. Goaillard JM, Taylor AL, Schulz DJ, Marder E: Functional consequences of
animal-to-animal variation in circuit parameters. Nat Neurosci 2009,
12:1424-1430.
3. Schulz DJ, Goaillard JM, Marder E: Variable channel expression in
identified single and electrically coupled neurons in different animals.
Nat Neurosci 2006, 9(3):356-362.
4. Taylor AL, Goaillard JM, Marder E: How Multiple Conductances Determine
Electrophysiological Properties in a Multicompartment Model. J Neurosci
2009, 29(17):5573-5586.
P174
Dynamical sensory representations establish a rapid odor code in a
spiking model of the insect olfactory system
Rinaldo Betkiewicz1,2*, Farzad Farkhooi1,2, Martin Paul Nawrot1,2,3
1
Theoretical Neuroscience / Neuroinformatics, Freie Universitt Berlin, Berlin,
Germany; 2Bernstein Center for Computational Neuroscience, Berlin,
Germany; 3Computational Systems Neuroscience, University of Cologne,
Cologne, Germany
E-mail: rinaldo.betkiewicz@bccn-berlin.de
BMC Neuroscience 2015, 16(Suppl 1):P174
In their natural environment, animals sense and evaluate olfactory cues of
time-varying composition and concentration. Their olfactory pathways are
adapted to the natural stimulus statistics, thus it is not surprising that odor
processing is fast [1]. Honey bees, for example, learn to discriminate odors
presented as short as 200 ms [2]. The neural odor code in these animals
emerges within 50ms after stimulus onset and neural representation
changes dynamically during and after an odorant is present [1,3]. How is
the insect olfactory system optimized to reliably estimate spatial and
temporal aspects of the olfactory environment and what are the
mechanisms behind rapid odor processing?
To investigate odor encoding at the Antennal Lobe (AL) and the
Mushroom Body (MB) level, we employ a simple phenomenological
spiking network model of the honeybee olfactory system. The model
implements a transformation from a low dimensional dense odorant
representation in the AL to a high dimensional sparse representation in
the MB. We demonstrate how information about the stimulus is present
in both encoding schemes, by time resolved classification of neural
activity.
Our model displays sparse and robust odor representation in the
Mushroom Body [4]. Typically, less than 10% of the Kenyon Cell population
is activated by an odor, with only 2-3 spikes at the odor onset (Figure 1A).
KC spikes establish a rapid odor identity code at stimulus onset, while
intrinsic adaptation currents provide a persistent and prolonged odor trace
(Figure 1B). Our approach allows us to investigate dynamical changes in
odor representations and predict odor after images.
Acknowledgements: BMBF grant 01GQ0941 Insect Inspired Robots
within the Bernstein Focus Learning and Memory (BFNL). Research
Figure 1(abstract P174) (A) Kenyon Cell spike raster plot.
Stimulation is indicated by gray shading (B) Top: Decoding Accuracy
given two odors (chance level: 0.5) as a function of time based on spike
count estimates in 50ms time bins. Bottom: Decoding accuracy based
on KC adaptation currents. Cellular adaptation levels provide a stable
odor trace that persists as an odor afterimage
Training Group Sensory Computation in Neural Systems (GRK 1589)
funded by the German DFG.
References
1. Nawrot MP: Dynamics of sensory processing in the dual olfactory
pathway of the honeybee. Apidologie 2012, 43:269-291.
2. Wright GA, Carlton M, Smith BH: A honeybees ability to learn, recognize,
and discriminate odors depends upon odor sampling time and
concentration. Behavioral Neuroscience 2009, 123(1):36-43.
3. Strube-Bloss M, Herrera-Valdez M, Smith B: Ensemble response in
mushroom body output neurons of the honey bee outpaces
spatiotemporal odor processing two synapses earlier in the antennal
lobe. PLoS One 2012, 7(11):e50322.
4. Farkhooi F, Froese A, Muller E, Menzel R, Nawrot MP: Cellular Adaptation
Facilitates Sparse and Reliable Coding in Sensory Pathways. PLoS
Computational Biology 2013, 9(10):e1003251.
P175
Application of generalized linear models to investigate functional
synaptic coupling and synchrony in an animal model of schizophrenia
Jennifer L Zick1,1*, Rachael K Blackman1,2,3, Matthew V Chafee1,3,
Theoden I Netoff1,4
1
Graduate Program in Neuroscience, University of Minnesota, Minneapolis,
MN 55455 USA; 2Medical Scientist Training Program (MD/PhD), University of
Minnesota, Minneapolis, MN 55455 USA; 3Brain Sciences Center, VA Medical
Center, Minneapolis, MN 55417 USA; 4Department of Biomedical
Engineering, University of Minnesota, Minneapolis, MN 55455 USA
E-mail: zick@umn.edu
BMC Neuroscience 2015, 16(Suppl 1):P175
Current hypotheses about the pathophysiology of schizophrenia suggest
that the disease results from disordered connectivity in the brain. Human
functional imaging studies have lent support to this idea by demonstrating
reduced temporal correlations between cortical areas in schizophrenic
patients [1]; however, functional evidence for disconnectivity at the
synaptic level is limited. Here we describe a study in which we applied
computational modeling techniques to assess functional connectivity
between spiking neurons in a pharmacological nonhuman primate model
of schizophrenia [2]. Neural data were obtained from multielectrode
recording arrays inserted into the parietal and prefrontal cortices of
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 119 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

macaque monkeys while the animals performed a cognitive control task

that measures a specific cognitive impairment in human patients with
schizophrenia. Phencyclidine, an NMDA receptor (NMDAR) antagonist with
well-described schizomimetic properties [3], was administered systemically
on alternating days with injections of saline.
To characterize changes in synaptic communication between neurons in
the disease state, we evaluated timing relationships in spike trains of
simultaneously recorded neurons. We employed a Generalized Linear
Model (GLM) approach [4] to infer patterns of functional connectivity
between neurons. The result of the GLM fit is a set of coupling functions
that estimate the change in the probability of a spike in one neuron in
response to a spike in another simultaneously recorded neuron. Unlike
traditional cross-correlation approaches used for inferring functional
connectivity, GLMs parse out the proportion of variance in spike timing
attributable to synaptic interactions between neurons from other sources
of variance in spike timing such as stimulus input, the intrinsic spike
patterns of the cells, and the effects of other simultaneously recorded
cells in the network [4]. After identifying functionally coupled pairs of
neurons using GLMs, we sought to evaluate the effects of phencyclidine
on functional coupling and synchrony between coupled cells. We found a
dramatic decrease in the proportion of cells that were functionally
coupled in the prefrontal cortex in the phencyclidine condition as
compared to the control condition. In order to investigate hypothesized
alterations in spike timing in the disease state, we identified the
distribution of interspike intervals between putative pre- and postsynaptic
neurons. In our analysis we found a prominent zero-lag peak
representing a large number of coincident action potentials between
coupled cells in the saline condition, but not in the phencyclidine
condition. This suggests a change in the timing of action potentials
which was apparent in the absence of a change in firing rates, precluding
overall decreased activity as an explanation for reduced synchrony. In
summary, these results suggest that phencyclidine induces a functional
disconnection between synaptically coupled cortical neurons which may
be related to a Hebbian reduction in synaptic strength resulting from
desynchronization of spiking activity. Figure 1(abstract P176) Circuit of the dis-inhibition model. The
Acknowledgements: This material is based upon work supported by the network was organized into 2 identical microcircuit units and consisted
NIH (R01 MH077779); Graduate Program in Neuroscience Training Award; of 4 types of neurons [4]: excitatory BOS neurons (triangles, EBO),
NSF Career Award (TIN); Medical Scientist Training Program NIH T32- excitatory grouping neurons (hexagon, G), feedforward inhibitory
008244. neurons (squares, FFI) and top-down inhibitory neurons (diamonds, TDI).
References The activity of a G neuron increased when a consistent figure was
1. Kim JJ, Kwon JS, Park HJ, Youn T, Kang DH, Kim MS, et al: Functional present in the receptive fields of EBO neuron pairs projecting to it [5],
disconnection between the prefrontal and parietal cortices during and it increased further when selective attention was directed to the
working memory processing in schizophrenia: A[15(O)]H2O PET study. object represented by this G neuron [6]
American Journal of Psychiatry 2003, 160(5):919-923.
2. Blackman RK, MacDonald AW, Chafee MV: Effects of Ketamine on Context-
Processing Performance in Monkeys: A New Animal Model of Cognitive
Deficits in Schizophrenia. Neuropsychopharmacology 2013, 38(11):2090-2100. (oblique lines) connect excitatory BOS (EBO) neurons representing one part
3. Morris BJ, Cochran SM, Pratt J: PCP. from pharmacology to modelling of a figure to feed-forward inhibitory (FFI) neurons representing another
schizophrenia. Curr Opin Pharmacol 2005, 5(1):101-106. figure part. Since horizontal connections are weakly myelinated or non-
4. Pillow JW, Shlens J, Paninski L, Sher A, Litke AM, Chichilnisky EJ, myelinated, signal propagation along them is slow and interactions
Simoncelli EP: Spatio-temporal correlations and visual signalling in a through them are subject to substantial delays, roughly proportional to the
complete neuronal population. Nature 2008, 454(7207):995-999. distance between the neurons they connect. We therefore investigated the
influence of communication delays on spike-spike synchrony between BOS
neurons. Our simulation results of the network with short delays are in
P176 agreement with experimental data, showing attentional enhancement of
The role of horizontal connections for the modulation of border- firing rates of EBO neurons and, at the same time, reduction of spike-spike
ownership selective neurons in visual cortex synchrony. In contrast, simulations with more realistic (longer) delays could
Nobuhiko Wagatsuma1*, Rudiger von der Heydt2, Ernst Niebur2 not reproduce experimental results. Our results suggest that horizontal
1
School of Science and Engineering, Tokyo Denki University, Hatoyama, Hiki, connections in early cortical areas cannot explain the observed synchrony
Saitama, Japan; 2Krieger Mind/Brain Institute, Johns Hopkins University, structure between BOS neurons and are unlikely to form the substrate of
Baltimore, MD, USA figure-ground segregation.
E-mail: nwagatsuma@rd.dendai.ac.jp Acknowledgements: Work partly supported by KAKENHI (no.26880019),
BMC Neuroscience 2015, 16(Suppl 1):P176 ONR (N000141010278), and NIH (R01EY016281-0). We are grateful to A
Martin for discussions.
Border-ownership selectivity (BOS) codes for the direction of a foreground References
object relative to the background [1]. Attention modulates the responses 1. Zhou H, Friedman HS, von der Heydt R: Coding of border ownership in
of BOS neurons in cortical areas V2, both for the firing rate [2] and in terms monkey visual cortex. J Neurosci 2000, 20:6594-6611.
of spike synchrony which can occur over long cortical distances [3]. Here, 2. Qiu FT, Sugihara T, von der Heydt R: Figure-ground mechanisms provide
we develop a network model of spiking neurons based on dis-inhibitory structure for selective attention. Nature Neurosci 2007, 10:1492-1499.
feedback [4]. Specifically, we consider the potential influence of intra-areal 3. Martin A, von der Heydt R: Firing synchrony between neurons reveals
("horizontal) connections for attention-induced modulation of BOS proto-object representation in monkey visual cortex. J Vision 2013,
neurons in visual cortex. In our model (Figure 1), horizontal connections 13:289.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
4. Buia CI, Tiesinga PH: Roles of interneuron diversity in the cortical
microcircuit for attention. J Neurophysiology 2008, 99:2158-2182.
5. Craft E, Schtze H, Niebur E, von der Heydt R: A neural model of figure-
ground organization. J Neurophysiology 2007, 97:4310-4326.
6. Mihalas S, Dong Y, von der Heydt R, Niebur E: Mechanisms of perceptual
organization provide auto-zoom and auto-localization for attention to
objects. PNAS 2011, 108(18):7583-7588.
P177
A computational model of cell culture dynamics: the role of
connectivity and synaptic receptors in the appearance of synchronized
bursting events
Davide Lonardoni*, Stefano Di Marco, Hayder Amin, Luca Berdondini,
Thierry Nieus
Istituto Italiano di Tecnologia, Genova, Italy
E-mail: davide.lonardoni@iit.it
BMC Neuroscience 2015, 16(Suppl 1):P177
How an ensemble of neurons wires together to form a functional unit is a
fundamental problem in neuroscience. The architecture of neuronal
wiring, in fact, determines how neurons communicate and may be
important for information processing performed by neuronal networks.
However current knowledge is mainly limited to networks consisting of a
small number of neurons, while the topological structure of biological
networks remains still unknown. Primary neuronal cultures represent an
ideal preparation to investigate the basic principles of network dynamics.
At the mature stage, they display network-wide synchronous bursting
events (SBEs) sharing similar spatio-temporal patterns of firing [1].
Interestingly, high-density MEA recordings have shown that SBEs actually
correspond to propagating activities through the network. Typically, the
simulated SBEs originated from a few and specific sites as in experiments
[2], but the nature and the role of such events is still under debate. In
order to investigate the determinants of such dynamics, we developed a
computational model that mimics the main features of the recordings
obtained by a high density multi-electrode-array device (4096 electrodes
inter-spaced by 20um, [3]). With only a few topological constraints, the
model expressed realistic SBEs along time that can be well clustered into
only a few groups differing for their ignition sites and propagation
directions, similarly to what it is observed experimentally. Furthermore,
we used the model together with experimental datasets to investigate
the effects of synaptic blockers of the AMPA, NMDA and GABA synapses
on the network activity. In particular, we showed that NMDA receptors
can be among the principal mechanisms involved in triggering a
sequence of SBEs, a firing regime that is typically observed in for mature
neuronal cultures. Such regime is characterized by a principal SBE
recruiting a great percentage of neurons and followed by a sequence of
several weaker SBEs interleaved by hundreds of milliseconds.
Altogether, the results obtained with our neural network computational
model show that this model can replicate most of the salient firing
properties observed experimentally in cultured neuronal networks and
that it can serve for exploring the properties of signals and responses
observed in neuronal networks properties.
Acknowledgements: We acknowledge the financial support of the SI-
CODE project of the Future and Emerging Technologies (FET) programme
within the Seventh Framework Programme for Research of The European
Commission, under FET-Open grant number: FP7-284553.
References
1. Raichman N, Ben-Jacob E: Identifying repeating motifs in the activation
of synchronized bursts in cultured neuronal networks. J Neurosci Methods
2008, 170(1):96-110.
2. Gandolfo M, Maccione A, Tedesco M, Martinoia S, Berdondini L: Tracking
burst patterns in hippocampal cultures with high-density CMOS-MEAs.
J Neural Eng 2010, 7(5):056001.
3. Berdondini L, Imfeld K, Maccione A, Tedesco M, Neukom S, Koudelka-
Hep M, Martinoia S: Active pixel sensor array for high spatio-temporal
resolution electrophysiological recordings from single cell to large scale
neuronal networks. Lab Chip 2009, 9(18):2644-2651.
Page 120 of 200
P178
PyRhO: a virtual optogenetics laboratory
Benjamin D Evans1*, Sarah Jarvis2, Simon R Schultz2, Konstantin Nikolic1
1
Institute of Biomedical Engineering, Department of Electrical & Electronic
Engineering, Imperial College London, London, UK; 2Department of
Bioengineering, Electrical Engineering, Imperial College London,
London, UK
E-mail: benjamin.evans@imperial.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P178
Optogenetics has become a key tool for understanding the function of
neural circuits and controlling their behaviour. An array of opsins have
been genetically isolated from several families of organism, including
algae and bacteria, with a wide range of temporal and spectral
properties. In an effort to develop more effective and tailored opsins,
hybrids and genetic mutants are continually being created.
Experimentally characterizing these new variants is a lengthy process
requiring substantial effort before they can be harnessed to address
questions in neuroscience. Experimentally testing each combination of
opsin and target cell type of interest is practically impossible, effectively
limiting the use of optogenetics as a tool. To aid in this effort we
propose PyRhO; an integrated suite of open-source, multi-scale
computational tools to characterize rhodopsins, then rapidly develop and
conduct virtual experiments with them in silico.
From a minimal set of photocurrent data, PyRhO will fit and
parameterize the Three [1], Four [1] and Six-state [2] rhodopsin models
to capture the underlying biophysical photocycle which defines their
kinetics. These models are then used to accurately compute the
photocurrents across a range of flux, voltage and other experimental
conditions for the given rhodopsin. After selecting a suitable model
based on the desired balance between simulation accuracy and speed,
the artificial rhodopsin can be seamlessly inserted into software such as
NEURON and Brian for use in simulations from the cellular to the
network level. We demonstrate the use of PyRhO in fitting models to
channelrhodopsin-2 (ChR2) [3] data and present results for typical
illumination strategies and experimental protocols designed to tease
apart the effects of key model parameters.
The tools are written in Python for easy scripting of experiments and
compatibility with a large array of open-source modules and software. An
accompanying GUI running in IPython [4] has also been developed to
facilitate more interactive exploration of the models for both
experimental and didactic purposes. Furthermore, IPython has been
identified as a particularly promising medium for sharing models and
reproducing results in computational neuroscience [5]. Simulations based
on these virtual opsins will enable neuroscientists to gain insight into
their behaviour and rapidly identify the most suitable variant for
application in a particular biological system, not only guiding choice, but
potentially also rhodopsin development. In this way, we expect PyRhO
will help to significantly improve the effectiveness of optogenetics as a
tool for transforming biological sciences.
Acknowledgements: This work was supported by the UK BBSRC grant
BB/L018268/1.
References
1. Nikolic K, Grossman N, Grubb MS, Burrone J, Toumazou C, Degenaar P:
Photocycles of Channelrhodopsin-2. Photochemistry and Photobiology
2009, 85:400-411.
2. Grossman N, Simiaki V, Martinet C, Toumazou C, Schultz SR, Nikolic K: The
spatial pattern of light determines the kinetics and modulates
backpropagation of optogenetic action potentials. Journal of
Computational Neuroscience 2013, 34(3):477-488.
3. Boyden ES, Zhang F, Bamberg E, Nagel G, Deisseroth K: Millisecond-
timescale, genetically targeted optical control of neural activity. Nature
Neuroscience 2005, 8(9):1263-1268.
4. Perez F, Granger BE: IPython: a System for Interactive Scientific
Computing. Computing in Science and Engineering 2007, 9(3):21-29.
5. Topalidou M, Leblois A, Boraud T, Rougier NP: A Long Journey into
Reproducible Computational Neuroscience. Frontiers in Computational
Neuroscience 2015, 9(30):1-2.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
P179
An efficient and accurate solver for large, sparse neural networks
Roman M Stolyarov1,2, Andrea K Barreiro1*, Scott Norris1
1
Department of Mathematics, Southern Methodist University, Dallas, TX, USA;
2
Harvard-MIT Department of Health Sciences and Technology, Cambridge,
MA, USA
E-mail: abarreiro@smu.edu
BMC Neuroscience 2015, 16(Suppl 1):P179
The mammalian brain has about 1011 neurons and 1014 synapses, with each
neuron presenting complex intra-cellular dynamics. The huge number of
structures and interactions underlying nervous system function thus make
modeling its behavior an extraordinary computational challenge. One
strategy to reduce computation time in networks is to replace
computationally expensive, stiff models for individual cells (such as the
Hodgkin-Huxley equations and other conductance-based models) with
integrate-and-fire models. Such models save time by not numerically
resolving neural behavior during its action potential; instead, they simply
detect the occurrence of an action potential, and propagate its effects to
postsynaptic targets appropriately. Thus, a complicated system of
continuous ordinary differential equations is replaced with a simpler, but
discontinuous, differential equation.
However, accurate existing methods for integrating discontinuous ordinary
differential equations (ODEs) scale poorly with problem size, requiring O
(N2) time steps for a system with N variables. The underlying challenge is
that discontinuities introduce O(dt) errors to conventional time integration
schemes, thus requiring very small time steps in the vicinity of a
discontinuity [1].
In this work, we propose a method to reduce this computational load by
embedding local network repairs within a global time-stepping scheme.
In addition, high-order accuracy can be achieved without requiring the
global time step to be bounded above by the minimum communication
delay, as is currently required in the hybrid time-driven/event-driven
scheme used by NEST [2]: this allows more powerful exploitation of exact
subthreshold [3,4] and quadrature-based [5] integration schemes. If the
underlying network is sufficiently sparse the algorithm, Adaptive Localized
Replay (ALR), will attain time complexity O(N) (Figure 1A). We apply our
method to a network of integrate-and-fire neurons that simulates
dynamics of a small patch of primary visual cortex (Figure 1B) [5,6].
Acknowledgements: This work was supported by the SMU Hamilton
Undergraduate Research Scholars Program (RS).
References
1. Shelley MJ, Tao L: Efficient and accurate time-stepping schemes for
integrate-and-fire neuronal networks. J Comp Neurosci 2001, 11(2):111-119.
Figure 1(abstract P179) (A) Comparison of runtime for a fully event-driven ("Full Replay) and ALR methods, for integrate-and-fire networks of
various system sizes N. (B) Raster plot of a 32 32 grid of V1 model neurons responding to a drifting grating stimulus. Inset: schematic of a subset of
the network, with selected synapses identified and shaded by strength. Red: AMPA; orange: NMDA, blue: fast GABA
Page 121 of 200
2. Gewaltig MO, Diesmann M: NEST (NEural Simulation Tool). Scholarpedia
2007, 2(4):1430.
3. Brette R: Exact simulation of integrate-and-fire models with synaptic
conductances. Neural Computation 2006, 18(8):2004-2027.
4. Morrison A, Straube S, Plesser HE, Diesmann M: Exact subthreshold
integration with continuous spike times in discrete-time neural network
simulations. Neural Computation 2007, 19(1):47-79.
5. Rangan AV, Cai D: Fast numerical methods for simulating large-scale
integrate-and-fire neuronal networks. J Comp Neurosci 2007, 22(1):81-100.
6. Cai D, Rangan AV, McLaughlin DW: Architectural and synaptic
mechanisms underlying coherent spiking activity in V1. Proceedings of
the National Academy of Sciences 2005, 102(16):5868-5873.
P180
Noise signature on interval timing
Sorinel A Oprisan*, Derek N Novo
Department of Physics and Astronomy, College of Charleston, Charleston, SC
29424, USA
E-mail: oprisans@cofc.edu
BMC Neuroscience 2015, 16(Suppl 1):P180
Among other essential adaptations is the capability of organisms to estimate
durations in the seconds-to-hours range (interval timing). Such capabilities
are critical for fundamental cognitive processes like decision making, rate
calculation, and planning of action [1]. In their seminal work on
computational modeling of interval timing, Matell and Meck [2] revitalized
the striatal beat frequency (SBF) model that utilizes the coincident activation
of a series of oscillators to code for different durations. They showed
through numerical simulations that the SBF model is capable of reproducing
two of the interval timing signatures: (a) precise timing, i.e. the model
output peaks at the training/criterion time and (b) scalar timing, i.e. the error
in timing increases linearly with the criterion time. The SBF model was
capable of producing scalar timing only when different types of biologically
realistic variances (frequency, memory, etc.) were considered [3].
In this work, we investigated what effect each type of variance/noise has on
the shape of the SBF model output. In particular, we noticed that the
experimentally measured response rate is not quite Gaussian (see Figure 1A)
and instead has a long tail. Mathematically, the output function of set of
coincidental (sinusoidal) oscillators is given by [3]:

out(t) = cos(2 fi T) cos(2 fi t),
where T is the criterion time, fi are the frequencies of neural oscillators.
The criterion time T is learned during the training phase and stored/
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P180) A. Normalized response for mice (squares - redrawn after [4]) versus normalized criterion time and the corresponding
best Gaussian fit (continuous line). B. Theoretical output from SBF model with different frequency variances
retrieved from the long-term memory with some errors. The firing
frequencies f i of all neural oscillators also fluctuate. We found that
memory variance (s T ) preserves the Gaussian shape of the output
function, whereas the frequency variance (sf) skews and has a long tail
similar to experimental observations (see Figure 1B).
In addition to the significantly difference contributions to the shape of
output functions, we also found that memory and frequency noises shift
the peak of the Gaussian. Memory noise affects the storage/retrieval of
criterion T and shifts the peak of the output to the right, i.e. tT = T(1+T),
where T is a number that depends on the range of stored values of T and
the probability distribution function (pdf T ) of the noise. Frequency
variance shifts the peak of the output to the left, i.e. tf = T/(1+f), where f
is a number that depended on the range of frequencies and the pdff of
the noise.
Acknowledgements: This work was supported by the grant IOS CAREER
1054914 from the National Science Foundation.
References
1. Gallistel CR: The organization of behavior Cambridge, MA: MIT Press 1990.
2. Matell MS, Meck WH: Cortico-striatal circuits and interval timing:
coincidence detection of oscillatory processes. Cognitive Brain Research
2004, 21(2):139-170.
3. Oprisan SA, Buhusi CV: What is all the noise about in interval timing?
Philosophical Transactions of the Royal Society of London B: Biological
Sciences 2014, 369(1637):20120459.
4. Buhusi CV, Aziz D, Winslow D, Carter RE, Swearingen JE, Buhusi MC: Interval
Timing Accuracy and Scalar Timing in C57BL/6 Mice. Behavioral
Neuroscience 2009, 123(5):1102.
P181
Spiking neural network model of reinforcement learning in the
honeybee implemented on the GPU
Esin Yavuz1*, Pascale Maul2, Thomas Nowotny1
1
CCNR, School of Engineering and Informatics, University of Sussex, Falmer,
Brighton, UK; 2Institute of Cognitive Science, University of Osnabrck, 49069
Osnabrck, Germany
E-mail: e.yavuz@sussex.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P181
Honeybees can learn and perform complex behavioral tasks despite their
small brains that contain less than a million neurons. At the same time
they are accessible to physiological experiments and the relatively small
number of neurons in their brain lends itself to quite detailed numerical
simulations. Bees therefore are a good model system for studying sensory
cognition and reinforcement learning.
We have shown in earlier work [1] that the anatomy and known
electrophysiological properties of the olfactory pathway of insects in
combination with spike-timing dependent plasticity (STDP) and lateral
inhibition lend themselves to an unsupervised self-organization of
synaptic connections for the recognition of odors. Here we extend this
Page 122 of 200
model by adding mechanisms of reinforcement learning, as suggested by
[2] (see Figure 1). We employ a three factor learning rule where plasticity
is governed by pre-synaptic and post-synaptic activity and a global
octopaminergic/dopaminergic reinforcement signal, triggered by a
reward. We investigated the role of feed-forward and feedback
mechanisms, as well as the role of the connectivity initially achieved by
unsupervised STDP.
Our model is implemented in the GeNN [3] framework, which facilitates
the use of GPUs for spiking neural network simulations using a code
generation framework. Because of the massive parallelism provided by
GPUs, we can simulate tens of thousands of neurons in real time in the
sparse firing regime relevant here. We investigated optimization strategies
and neuron and synapse model choices for a better performance on the
GPU. The model presented here is a stepping-stone to more sophisticated
learning models and multi-sensory integration in the Green Brain Project
[4], in which we aim to control a flying robot with a simulation of learning
and decision making mechanisms in the honeybee related both to the
olfactory and visual pathways.
Acknowledgements: This project is supported by the EPSRC (Green
Brain Project, grant number EP/J019690/1).
References
1. Nowotny T, Huerta R, Abarbanel HD, Rabinovich MI: Self-organization in
the olfactory system: one shot odor recognition in insects. Biological
Cybernetics 2005, 93(6):436-446.
2. Smith D, Wessnitzer J, Webb B: A model of associative learning in the
mushroom body. Biological Cybernetics 2008, 99(2):89-103.
3. Yavuz E, Turner J, Nowotny T: Simulating spiking neural networks on
massively parallel graphical processing units using a code generation
approach with GeNN. BMC Neuroscience 2014, 15(Suppl 1):O1.
4. The Green Brain project. [http://www.greenbrainproject.co.uk/], accessed
20-02-2015.
P182
Estimating numerical error in neural network simulations on Graphics
Processing Units
James P Turner*, Thomas Nowotny
Centre for Computational Neuroscience and Robotics, University of Sussex,
Brighton, UK
E-mail: J.P.Turner@sussex.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P182
Modern graphics processing units (GPUs) are becoming a popular
hardware substrate for spiking neural network simulations [1-3], due to
their massive parallelism and impressive cost-to-speed ratio. However,
verifying and interpreting the results of a GPU simulation can be difficult
because the results are never exactly reproducible, unlike those from an
equivalent serial simulation on a CPU; not only is the simulation subject
to the usual rounding errors of floating-point arithmetic, but there are
also elements of stochasticity due to the non-determinism of the thread
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 123 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

Figure 1(abstract P181) Network diagram for the hypothesized model of reinforcement learning in the honeybee olfactory system. Excitatory
connections are shown in black, inhibitory connections in blue and learning synapses in red. Grey arrows represent the abstractions modeled by implicit
mechanisms. The model consists of the antennal lobe, lateral horn interneurons, mushroom body Kenyon cells and lobe neurons, and an octopaminergic/
dopaminergic pathway for reinforcement, classically considered to be the VUMmx1 neuron. A conditioned stimulus is paired with an unconditioned
stimulus (sugar to the antenna) to elicit the behavioral response (proboscis extension) in the training phase, which can then be rewarded by letting the
bees drink. The association with reward facilitates plasticity in the synapses between the Kenyon cells and lobe neurons and between lobe neurons and
pre-motor neurons. The size of weight changes is determined by an eligibility trace as a function of the delay between stimulus and reward

scheduling mechanism on the hardware, alongside the non-associativity simulations. Furthermore, both CPU and GPU implementations are
of floating-point addition and multiplication. Consider, for example, a compared against an equivalent simulation using an accurate arbitrary-
typical postsynaptic integration step for the summation of incoming precision floating-point arithmetic library, to determine how far the CPU and
synapse currents, executed on a GPU. If there are multiple threads, which GPU simulation trajectories deviate from the analytically correct trajectory.
are each simulating an incoming synapse, there is no guarantee for the For illustration, the divergence of a single neuron in a 10,000 neuron
order in which each synapse threads current will be accumulated into
the total current. Therefore, rounding errors will be different depending
on this order and the result could be different every time the simulation
runs. Such effects would initially be small but can be amplified in
unstable or chaotic systems to a degree that the final results appear
completely random across different runs (see Figure 1).
When comparing runs between GPU and CPU implementations there are
additional sources of divergence. There are subtle differences in the way
each architecture implements floating-point arithmetic. For instance, the
NVIDIA C2070 GPU tested in this study implements the fused multiply-add
(FMA) operation, introduced in the latest IEEE-754-2008 floating-point
standard, whereas most Intel CPUs perform the multiplication and addition
operations separately, with lower accuracy. Only Intels most recent Haswell
CPU architecture implements the more accurate FMA operation, whilst
many current lab workstations contain chips that do not.
The aim of the work presented here is to analytically determine the
theoretical worst-case and average-case numerical absolute error incurred Figure 1(abstract P182) In repeated runs, results of numerical
when simulating neural network models on an NVIDIA CUDA GPU. These simulations on GPUs can vary. Mean, standard deviation and range of
error measurements are also compared with the absolute error resulting observed membrane potential of a neuron in a network of 10,000
from the equivalent serial algorithm running on a single CPU core, using Izhikevich neurons, 8,000 excitatory and 2,000 inhibitory, with 1,000
standard float32 (float) and float64 (double) precision floating-point random connections each; after only 190 ms simulation the results start
arithmetic, to determine a reasonable error margin for verifying the results to diverge visibly, and after only 210 ms, they differ largely
of parallel GPU simulations against those of equivalent serial CPU
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Izhikevich network is plotted in the figure. Finally, we also analyse the role of
errors originating from approximate integration methods and compare them
to the underlying numerical errors discussed thus far.
References
1. GeNN. [https://github.com/genn-team/genn], accessed 20-02-2015.
2. NeMo. [http://nemosim.sourceforge.net/], accessed 20-02-2015.
3. CARLsim. [http://www.socsci.uci.edu/~jkrichma/CARLsim/], accessed 20-02-
2015.
P183
Low-dimensional spike rate dynamics of coupled adaptive model neurons
Moritz Augustin1,2*, Josef Ladenbauer1,2, Klaus Obermayer1,2
1
Neural Information Processing Group, Berlin Institute of Technology, Berlin,
Germany; 2Bernstein Center for Computational Neuroscience Berlin, Berlin,
Germany
E-mail: augustin@ni.tu-berlin.de
BMC Neuroscience 2015, 16(Suppl 1):P183
The spiking activity of single neurons can be well described by a two-
dimensional integrate-and-fire model that includes neuronal adaptation [1]
caused by slowly decaying potassium currents [2]. For fluctuating inputs
sparsely coupled spiking model neurons exhibit stochastic population
dynamics which can be effectively characterized using the Fokker-Planck
equation (see, e.g., [3-5]). This approach leads to a model with an infinite-
dimensional state space and non-standard boundary conditions. However,
the spike rate dynamics can be approximated by a low-dimensional
ordinary differential equation in different ways [4,6,7]. Although these
approximation techniques are interrelated it is not clear which reduced
model best reproduces the spike rate of the original spiking network,
depending on the statistics of the input. Here we first extend each of these
reduction methods to account for neuronal adaptation and then evaluate
the reduced models in terms of spike rate reproduction accuracy for a
range of biologically plausible input statistics, computational demand and
implementation complexity (see, e.g., Figure 1). These reduced descriptions
are well suited for (i) application in neural mass/mean-field based brain
network models, having a link to single neuron properties retained and
being computationally efficient, and (ii) mathematical analyses of, e.g.,
asynchronous and rhythmic network states.
Acknowledgements: This work was supported by the DFG Collaborative
Research Center SFB910.
References
1. Brette R, Gerstner W: Adaptive exponential integrate-and-fire model as an
effective description of neuronal activity. J Neurophysiol 2005,
94:3637-3642.
2. Ladenbauer J, Augustin M, Obermayer K: How adaptation currents change
threshold, gain, and variability of neuronal spiking. J Neurophysiol 2014,
111:939-953.
Figure 1(abstract P183) Simulation of a large population of
adaptive exponential integrate-and-fire (aEIF) neurons driven by a
stochastic current with time-varying moments. Instantaneous spike
rate and adaptation current averaged over 200,000 neurons are shown
in gray. Overlaid are mean spike rate and adaptation current of two
derived low-dimensional models receiving input with the same time-
dependent moments as the population of aEIF neurons
Page 124 of 200
3. Brunel N: Dynamics of sparsely connected networks of excitatory and
inhibitory spiking neurons. J Comput Neurosci 2000, 8:183-208.
4. Mattia M, Del Giudice P: Population dynamics of interacting spiking
neurons. Phys Rev E 2002, 66:051917.
5. Augustin M, Ladenbauer J, Obermayer K: How adaptation shapes spike
rate oscillations in recurrent neuronal networks. Front Comput Neurosci
2013, 7:9.
6. Schaffer E, Ostojic S, Abbott L: A complex-valued firing-rate model that
approximates the dynamics of spiking networks. PLOS Comput Biol 2013,
9:e1003301.
7. Ostojic S, Brunel N: From spiking neuron models to linear-nonlinear
models. PLOS Comput Biol 2011, 7:e1001056.
P184
pypet: a python toolkit for simulations and numerical experiments
Robert Meyer1,2*, Klaus Obermayer1,2
1
Neuroinformatics Group, Technische Universitaet Berlin, 10587 Berlin,
Germany; 2Bernstein Center for Computational Neuroscience Berlin, 10115
Berlin, Germany
E-mail: robert.meyer@ni.tu-berlin.de
BMC Neuroscience 2015, 16(Suppl 1):P184
pypet (python parameter exploration toolkit [1]) is a new multi-platform
python toolkit for management of simulations and storage of numerical
data. Exploring or sampling the space of model parameters is one key
aspect of simulations and numerical experiments. pypet was especially
designed to allow easy and arbitrary sampling of trajectories through a
parameter space beyond simple grid searches. Moreover, special focus is
put on managing different neuron models in python network simulations
like BRIAN [2]. Simulation parameters as well as the obtained results are
collected by pypet and stored in the widely used HDF5 file format [3].
This allows fast and convenient loading of data for further analyses.
Furthermore, pypet provides an environment with various features. For
example, among these are multiprocessing for fast parallel simulations,
dynamic loading of data, integration of Git version control, and supervision
of experiments via the electronic lab notebook Sumatra [4]. A rich set of
data formats is supported, encompassing native python types, numpy and
scipy data, pandas DataFrames [5], and data from BRIAN [2]. Moreover, the
toolkit is easily extendible to allow the user to add customized data
formats. pypet is a very flexible tool and suited for short python scripts as
well as large scale projects. Thus, pypet supports reproducible research in
computational neuroscience and other disciplines that involve simulations
and numerical experiments.
References
1. Meyer R, Obermayer K: pypet [http://pypet.readthedocs.org/], online
documentation, 2013-2015.
2. Goodman DF, Brette R: The Brian simulator. Front Neurosci 2009,
3(2):192-197[http://briansimulator.org/].
3. The HDF Group: Hierarchical Data Format, version 5. 1997-2015 [http://www.
hdfgroup.org/HDF5/].
4. Davison AP: Automated capture of experiment context for easier
reproducibility in computational research. Computing in Science and
Engineering 2012, 14(4):48-56[http://neuralensemble.org/sumatra/].
5. McKinney W: Data Structures for Statistical Computing in Python.
Proceedings of the 9th Python in Science Conference 2010, 51-56[http://
pandas.pydata.org].
P185
Extending integrate-and fire model neurons to account for the effects
of weak electric fields in the presence of dendrites
Florian Aspart1,2*, Josef Ladenbauer1,2, Klaus Obermayer1,2
1
Neural Information Processing Group, Berlin Institute of Technology, Berlin,
Germany; 2Bernstein Center for Computational Neuroscience Berlin, Berlin,
Germany
E-mail: florian.aspart@ni.tu-berlin.de
BMC Neuroscience 2015, 16(Suppl 1):P185
Transcranial brain stimulation techniques have recently sparked a strong
interest in understanding the effects of weak electric fields on neuronal
network dynamics (e.g. [1,2]). The collective dynamics of large populations
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
of coupled neurons can be efficiently studied using single-compartment
(point) model neurons of the integrate-and-fire (IF) type [3], which allow
for a systematic model reduction at the population level [4,5], as opposed
to multi-compartment Hodgkin-Huxley type models and complex
morphologies. However, existing point neuron models cannot adequately
reproduce the effects of an electric field on the somatic membrane
potential, which are influenced by the presence of dendritic processes [2].
Here, we present an extension for IF type point neuron models to take into
account the subthreshold effects of an oscillating weak uniform
extracellular field, similar to those generated in the brain by transcranial
electrical stimulation [6]. Based on a ball-and-stick neuron model (i.e., a
passive finite dendritic cable with a lumped soma at its end) we
analytically calculate the somatic membrane polarization induced by a
weak extracellular electric field using the cable equation. From this
polarization we derive an equivalent input current for leaky IF as well as
adaptive nonlinear IF point neurons, which explicitly depends on the
(soma+dendrite) neuron model and electric field parameters. The
extended point neuron model can well reproduce the relationships
between electric field properties (intensity, frequency) and neuronal
responses (membrane polarization, sensitivity and phase), as observed by
simulations of neuron models with complex morphologies and reported in
the experimental literature [7]. Our point neuron model extension is simple
to implement and well suited for application in IF based neural networks.
Acknowledgements: This work was supported by the DFG Priority
Program SPP1665 and DFG Collaborative Research Center SFB910
References
1. Ali MM, Sellers KK, Frhlich F: Transcranial alternating current stimulation
modulates large-scale cortical network activity by network resonance.
J Neurosci 2013, 33(27):11262-11275.
2. Reato D, Rahman A, Bikson M, Parra LC: Effects of weak transcranial
alternating current stimulation on brain activity-a review of known
mechanisms from animal studies. Front Hum Neurosci 2013, 7:687.
3. Reato D, Rahman A, Bikson M, Parra LC: Low-Intensity Electrical
Stimulation Affects Network Dynamics by Modulating Population Rate
and Spike Timing. J Neurosci 2010, 30(45):15067-15079.
4. Augustin M, Ladenbauer J, Obermayer K: How adaptation shapes spike
rate oscillations in recurrent neuronal networks. Front Comput Neurosci. 2
013, 7:9.
5. Ladenbauer J, Augustin M, Obermayer K: How adaptation currents change
threshold, gain, and variability of neuronal spiking. J Neurophysiol 2014,
111:939-953.
6. Datta A, Bansal V, Diaz J, Patel J, Reato D, Bikson M: Gyri-precise head
model of transcranial direct current stimulation: improved spatial
focality using a ring electrode versus conventional rectangular pad. Brain
Stimul 2009, 2(4):201-207.
7. Deans JK, Powell AD, Jefferys JGR: Sensitivity of coherent oscillations in
rat hippocampus to AC electric fields. J Physiol 2007, 583(Pt 2):555-565.
P186
Decoding of naturalistic textures from spike patterns of neuromorphic
artificial mechanoreceptors
Alberto Mazzoni*, Udaya B Rongala, Calogero M Oddo
The BioRobotics Institute, Scuola Superiore SantAnna, Viale Rinaldo Piaggio
34, Pontedera 56025, Pisa, Italy
E-mail: a.mazzoni@sssup.it
BMC Neuroscience 2015, 16(Suppl 1):P186
Significant advances were recently achieved in the bidirectional control of
upper limb neuroprostheses [1], however implementing a realistic tactile
feedback on artificial hands is still an open challenge [2]. We target this
ambition with a neuromorphic approach. Towards this direction we
integrated MEMS tactile sensors with a realistic spatial arrangement on the
distal phalanx of artificial fingers [3] and we converted the sensor readouts
into spike trains mimicking the neural firing properties of glabrous skin
mechanoreceptors [4]. The neuromorphic approach presents several
advantages [5], among which the most relevant for neuroprosthetics is
that a sufficiently realistic tactile feedback is expected to reduce drastically
the time the brain needs to adapt to the prosthesis and to each single new
task, leading to enhanced quality of life of the subjects.
Since the long term goal is to implement a hardware version of the
neuromorphic neurons, we modeled the mechanoreceptors as Izhikevich
Page 125 of 200
neurons, a model which at the same time is computationally lean and
takes into account the effects of firing rate adaptation that are
particularly relevant in mechanoreceptors. Parameters were tuned so to
achieve a match with Slowly Adapting (SA) mechanoreceptors dynamics
in primates [4]. The output of each tactile sensor was normalized and
injected as an excitatory input current into a single neuron model [6]. To
reproduce the features of Fast Adapting (FA) mechanoreceptors
responsible for edge detection [7], we injected in a second set of neuron
models the smoothed derivative of the pressure sensors outputs.
The fingertip was then repeatedly presented with ten different daily-life
textures such as textiles, polymeric tissues, glass or paper. FA spike trains
were used to identify onset and offset of contact between fingertip and
stimuli, whereas SA spike trains were used to decode the stimulus
presented. We applied two decoding procedures: one based on the
combination of firing rate and inter-spike interval Fano factor, and the
other one based on Victor-Purpura distance [8]. Error rate was close to 20%
in the first case and to 5% in the second, reflecting the relevance for
decoding of the fine temporal structure of the neuromorphic sensor
outputs. Finally, we introduced a normalization in the cost of the Victor-
Purpura spike train metrics thanks to which the above level of correct
classification was maintained even if sensing conditions varied across trials.
These results support a possible future evaluation of the biomimetic
fingertip endowed with neuromorphic artificial mechanoreceptors in
clinical trials with human subjects.
Acknowledgements: This work was supported by the EU Grant FP7-FET
611687 NEBIAS project, by the EU Grant FP7-NMP 228844 NANOBIOTOUCH
project, by the Italian PRIN HandBot project, and by Scuola Superiore
SantAnna.
References
1. Raspopovic S, Capogrosso M, Petrini FM, Bonizzato M, Rigosa J, Di Pino G,
et al: Restoring Natural Sensory Feedback in Real-Time Bidirectional
Hand Prostheses. Science Translational Medicine 2014, 6(222):222ra219.
2. Kwok R: Neuroprosthetics: once more, with feeling. Nature 2013,
497(7448):176-177.
3. Oddo CM, Controzzi M, Beccai L, Cipriani C, Carrozza MC: Roughness
Encoding for Discrimination of Surfaces in Artificial Active-Touch. IEEE
Transactions on Robotics 2011, 27(3):522-533.
4. Johansson RS, Flanagan JR: Coding and use of tactile signals from the
fingertips in object manipulation tasks. Nat Rev Neurosci 2009,
10(5):345-59.
5. Service RF: Minds of their own. Science 2014, 346(6206):182-183.
6. Izhikevich EM: Simple model of spiking neurons. IEEE Transactions on
Neural Networks 2003, 14(6):1569-1572.
7. Spigler G, Oddo CM, Carrozza MC: Soft-neuromorphic artificial touch for
applications in neuro-robotics. 2012 4th Ieee Ras & Embs International
Conference on Biomedical Robotics and Biomechatronics (Biorob) 2012,
1913-1918.
8. Victor JD, Purpura KP: Nature and precision of temporal coding in visual
cortex: A metric-space analysis. Journal of Neurophysiology 1996,
76(2):1310-1326.
P187
Hierarchical organization of multiscale communities in brain networks
is non-tree structured
Hiroshi Okamoto1,2
1
RIKEN Brain Science Institute, Saitama, 351-0198, Japan; 2Research &
Development Group, Fuji Xerox Co. Ltd., Kanagawa, 220-8668, Japan
E-mail: hiroshi.okamoto@fujixerox.co.jp
BMC Neuroscience 2015, 16(Suppl 1):P187
In literature of network science, a group of nodes that are densely
connected within the group and are less connected with nodes outside
the group is referred to as a community [1]. Community structure is a
fundamental property of a variety of social, biological and engineering
networks. Specifically, communities in brain networks are considered to be
associated with functional modules of information processing in the brain
[2]. To reveal information processing architecture of the brain, therefore, it
is pivotal to know individual communities and their organization in brain
networks.
Community structure in brain networks is characterized by hierarchical
organization, which reflects that functional modules at larger scales are
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 126 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

Figure 1(abstract P187) Hierarchical organization of communities in the neuronal network of C. elegans. Each layer of the hierarchy is shown by
horizontal alignment of red squares. Each square indicates a community detected at each layer. The size of each square indicates the size of the
corresponding community. Note that many communities have more than one link from the upper layer, which means that these communities have more
than one parent. These demonstrate non-tree structure of hierarchical organization of communities in the brain network

built up from a set of functional modules at smaller scales [3]. A number
of mathematical methods for detecting communities in networks have
been developed so far [1], but unfortunately few of them can consistently
deal with hierarchical organization of multiscale communities. Here we
propose a reliable method for detecting hierarchical organization of
multiscale communities. Then we examine community structure of real
brain networks by use of this method.
The proposed method is based on a novel Bayesian formulation of
Markov chain. The method has only one parameter, , which comes from
the precision of the prior distribution of a random process. The amplitude
of controls the resolution of community detection; the smaller its
amplitude, the finer the size of detected communities. Quasi-static
increase in causes a series of discrete phase transitions; at each transition
point a subset of smaller communities (children) agglomerate a larger
community (parent), thus leading to a hierarchical organization of
multiscale communities.
Applying this method to the neuronal network of C. elegans [4] and the
macaque cortical network [5], we have found that hierarchical organization
of multiscale communities in these networks is non-tree structured: Some
child communities have more than one parent community (Figure 1).
These findings suggest efficient architecture for integration of functional
modules in brain information processing: The same functional modules at
lower levels can be shared by distinct functional modules at higher levels.
Acknowledgements: This work was partly supported by JSPS KAKENHI
Grant Number 23500379.
References
1. Newman MEJ: Communities, modules and large-scale structure in
networks. Nature Phys 2012, 8:25-31.
2. Bullmore E, Sporns O: Complex brain networks: graph theoretical analysis
of structural and functional systems. Nature Rev Neurosci 2009,
10(3):186-198.
3. Meunier D, Lambiotte R, Bullmore ET : Modular and hierarchically modular
organization of brain networks. Front Neurosci 2010, 4:1-11.
4. Watts DJ, Strogatz SH: Collective dynamics of small-world networks.
Nature 1998, 393:440-442.
5. CoCoMac. [http://cocomac.g-node.org/].
P188
Sourcing brain histone modification data and development of
algorithm for identification of hypersensitive sites
Victor Chukwudi Osamor1,2
1
Department of Computer and Information Sciences, Covenant University, P.
M.B 1023, Ota, Ogun State, Nigeria; 2Institute of Informatics, University of
Warsaw, ul Banacha 2, 02-097, Warsaw, Poland
E-mail: vcosamor@gmail.com
BMC Neuroscience 2015, 16(Suppl 1):P188
Background: The source of data for computational analysis goes a long
way to determine the quality of data used and the computed result. This
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
is due to the differences in the avalanche of protocols applicable in
experimental data generation, professional skills, endurance, diligence
and consideration to details by investigators. The relevance of
identification of DNase hypersensitive sites [1] gives a clue to the role of
genes based on the transcription binding properties of various regions.
The aim of this study is to conduct a comparative assessment of the
ready availability of brain histone modification data and propose an
algorithm for the use of such data for transcription factor dimer
prediction.
Methods: We carried out an extensive analysis on Encyclopedia of DNA
Elements (ENCODE) [2] to determine the easiest way of accessing histone
modification (HM) data [3]. We compared the process of entering the
web address of http://genome.ucsc.edu/ to invoke the genome browser
by a click Select the human genome and click submitScrolling
downwards till Regulation which is a blue horizontal band finding
and clicking ENCODE Histone Modification Selecting the source of
desired histone among four different sources and finally downloading
from huge array of scrolling pages of data to locate your data of interest.
This is quite cumbersome compared to a second method goggling
ENCODE Data Matrix and select ChIP-seq option to see all the data
content arranged neatly in form of a matrix. Comparatively, we
considered 12 HMs in all for both Tier 1 and brain cell lines. Using Tier 1
(GM12878, H1hESC, K562) cell lines, we compared their availability with
H1-Neurons and other brain cell-type viz Glioba, Medullo NH-A, Hac, Be2c
and Sknshra as depicted in Table 1 with symbol of + representing
present and - representing absent.
Results: The HM of cell line in Tier 1 cell type is available for almost all
considered available HMs while that of brain cells and H1-neurons were a
either totally absent or scarcely available. In addition, Nh-A appears to be
the only exceptional case where the availability is equal to the Tier 1
histone modification. We there evolved an algorithm that identifies
hypersensitive sites from these data in the following way: 1. Run HM bam
files on Model-based Analysis of ChIP-Seq(MACS) using non-modal, no
lambda [4] parameter setting 2. If single HM, no merge is required then
go to item 4. 3. If merge is required for a combination of HM peaks then
(a) Append (b) Sort and (c) apply merge option 4. Prepare the various
threshold to be clustered by a dimer prediction algorithm [5].
Conclusions: A deluge of data can sometimes be bewildering especially
when searching for a cell line of interest in a large database but we can
arrest the difficulties following the proposed technique. The scarce
presence of HM for brain cells calls for more attention on for the choice
of such cell type for further investigation.
Acknowledgements: The research leading to these results has received
funding from the European Union Seventh Framework Programme (FP7/
2007-2013) under grant agreement n 246016. This work was done at
University of Warsaw
References
1. Koohy H, Down TA, Hubbard TJ: Chromatin accessibility data sets show
bias due to sequence specificity of the DNase I enzyme. PLoS One 2013,
8(7):e69853.
2. The ENCODE Project Consortium: An Integrated Encyclopedia of DNA
Elements in the Human Genome. Nature 2012, 489(7414):57-74.
3. Thurman RE, Rynes E, Humbert R, Vierstra J, Maurano MT, Haugen E,
Sheffield NC, et al: The Accessible Chromatin Landscape of the Human
Genome. Nature 2012, 489(7814):75-82.
4. Zhang Y, Liu T, Meyer CA, Eeckhoute J, Johnson DS, Bernstein BE, et al:
Model-based Analysis of ChIP-Seq(MACS). Genome Biology 2008, 9(9):R137.
5. Jankowski A, Prabhakar S, Tiuryn J: TACO: a general-purpose tool for
predicting cell-type-specific transcription factor dimers. BMC Genomics
2014, 15:208-219.
P189
Vibrational resonance in feed-forward-loop neuronal network motifs
Ali Calim1*, Ugur Ileri1, Muhammet Uzuntarla1, Mahmut Ozer2
1
Department of Biomedical Engineering, Bulent Ecevit University, Zonguldak
67100, Turkey; 2Department of Electrical and Electronics Engineering, Bulent
Ecevit University, Zonguldak 67100, Turkey
E-mail: ali.calim@beun.edu.tr
BMC Neuroscience 2015, 16(Suppl 1):P189
Page 127 of 200
A neuron carries out its functions in networks receiving contacts from
roughly 104 presynaptic neurons. Such a dense connectivity profile for a
single neuron may give rise to enormous complex neuronal topologies,
which might be very difficult to understand the underlying mechanisms
of neurological functions and diseases. Extensive experimental data from
neuroanatomical studies have uncovered that neural networks include
some recurring topologies of microcircuits, known as network motifs
which serve as characteristic building blocks of complex networks [1,2].
Therefore, it is widely assumed that a clear explanation on dynamical and
functional features of these network motifs can be considered as the first
step to understand large networks.
Following this motivation, in this study, we investigate the Vibrational
Resonance (VR) phenomenon in a triple-neuron feed-forward-loop (FFL)
which is one of the most significant brain network motifs, shown in
Figure 1. VR is a physical phenomenon found in nonlinear systems
where a weak signal can be detected and processed by the system
with the assistance of another high frequency signal. It is very similar to
the well-known stochastic resonance phenomenon, where the role of
noise is replaced in VR by a high frequency signal. In recent years,
there exists a growing interest in applications of VR to neuroscience
because two frequency signals are pervasive in neural systems, i.e.
bursting neurons exhibit two widely different time scales, simultaneous
arrival of vocal signals having distinct frequencies to auditory neurons.
Due to the significance of its potential in neural signal processing, the
VR phenomenon has already been studied with computational models
of neurons and their networks [3,4]. However, a population of excitatory
and inhibitory neurons has not been considered, yet. Here, we study
the VR dynamics depending on whether the neurons in the FFL motifs
are excitatory or inhibitory considering eight possible structural
configurations of the considered microcircuit (Table 1). Such an
approach provides us to test the functional influences of structural
configuration on VR dynamics in neural populations.
We found that the weak signal transmission in the network via the VR
mechanism is possible depending on both the coupling strength between
neurons and the network structure. For instance, when a weak coupling
strength is present between connected neurons, weak signal injected to
the input neuron can be transmitted to the output with only T8-FFL motif.
In this case, signal transmission is impossible for other types of network
motifs. For intermediate coupling strengths, signal transmission
performance is high for all motifs and motif type does not change very
much VR dynamics. Finally, for strong coupling strengths, the best VR
performance is obtained with T1-FFL motif where all the neurons in the
network are excitatory. We also clarify the mechanisms that underlies the
differences in performance of network motifs.
References
1. Sporns O, Ktter R: Motifs in brain networks. PLoS Biol 2004, 2(11):369.
2. Guo D, Li C: Stochastic and coherence resonance in feed-forward-loop
neuronal network motifs. Phys Rev E 2009, 79(5):051921.
3. Deng B, Wang J, Wei X, Tsang KM, Chan WL: Vibrational resonance in
neuron populations. Chaos 2010, 20:013113.
4. Uzuntarla M, Yilmaz E, Wagemakers A, Ozer M: Vibrational resonance in a
heterogeneous scale free network of neurons. Commun Nonlinear Sci
Numer Simul 2015, 22(1-3):367-374.
P190
Neural representation in F5: cross-decoding from observation to
execution
Murat Kirtay1, Vassilis Papadourakis2,3, Vassilis Raos2,3, Erhan Oztop1*
1
Computer Science, Ozyegin University, Istanbul, Turkey; 2Foundation for
Research & Technology- Hellas (FORTH) , Heraklion, Greece; 3University of
Crete Medical School, Heraklion, Greece
E-mail: erhan.oztop@ozyegin.edu.tr
BMC Neuroscience 2015, 16(Suppl 1):P190
Mirror neurons fire during both action execution and observation of a
similar action performed by another individual [1]. However, this definition
does not account for the existence of representational equivalence
between execution and observation. To investigate this issue we recorded
68 neurons from area F5 of a macaque monkey trained either to execute
reaching-to-grasp actions towards objects or to observe the experimenter
performing the same actions [2], and adopted a decoding framework to
 BMC Neuroscience 2015, Volume 16 Suppl 1

Table 1(abstract P188) Comparative histone modification analysis of Tier 1 (GM12878, H1hESC, K562) with others consisting of brain cell line and neuron
H2AFZ H3K27ac H3K27me3 H3K36me3 H3K4me1 H3K4me2 H3K4me3 H3K79me2 H3K9ac H3K9me1 H3K9me3 H4K20me1
GM12878 + + + + + + + + + - + + 91.67%
HI-hESC + + + + + + + + + - + + 91.67%
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

K562 + + + + + + + + + + + + 100.00%
H1-Neurons - - - - - - - - - - - - 0%
Glioba - - - - - - - - - - - - 0%
Medullo - - - - - - - - - - - - 0%
NH-A + + + + + + + + + - + + 91.67%
Hac - - - - - - + - - - - - 0%
Be2c - - - - - - + - - - - - 8.33%
Sknshra - - + + - - + - - - - - 25%
Symbol + indicates present and - indicates absent
Page 128 of 200
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P189) Schematic illustration of the considered
feed-forward-loop motif. LF and HF refers to low and high frequency
signals, respectively. Neuron 1 and 3 is considered as input and output
neurons, respectively
find whether neurons effective in decoding the object/grip type in (1)
execution and (2) observation conditions do exist, and most critically, to (3)
assess whether transfer between execution and observation decoders (i.e.
cross-decoding) can be employed. By transfer we mean the application of
the decoder parameters estimated using the neural discharge in observation
to the neural firing recorded in execution, and vice versa. The success rate of
such a decoder indicates the equivalence of representations in the two
conditions.
Our analysis indicates that, at the level of single neurons, object/grip-specific
decoders can be constructed, i.e. the type of the object/grip employed in
either execution or observation can be decoded (success rate: 80%-100%,
chance level: 25%). However, only in 10% of the cases (corresponding to the
congruent type mirror neurons [1]) the decoder based on the execution
discharge was effective when transferred to the observation discharge. The
same was true for the reverse transfer. To extend this analysis at the
population level we examined all pair performance of a 10-neuron set,
consisted of 4 neurons having the best decoding performance and
6 neurons randomly selected. Out of the 45 possible pairs, 7 displayed high
success rates (80% on average) in cross-decoding. Remarkably, high
performing pairs were constituted only when one of the neurons displaying
reliable decoding performance was paired with a randomly selected -poor
solo decoder- neuron, which acted as a helper. These results strongly point
to a population based representation where good and poor decoders may
cooperate to form a robust recognition system.
Methods: Neuronal discharges during each condition were trimmed and
represented as 14-bin histogram vectors. In the two neuron analysis,
each neuron was reduced to a 7-bin histogram, and their concatenation
results in a 14-tuple vector, to ensure similar decoder complexity
(constant number of adjustable parameters). Thus, for each condition, ten
14-tuple neural firings (one per trial) made up the rows of the input
matrix X, and the corresponding object ids (1-4) made up the output
Table 1(abstract P189) Eight possible FFL types
Type Neuron 1 Neuron 2 Neuron 3
T1-FFL E E E Conclusions: Our result shows the multi-modal features of bistable
T2-FFL E I E
T3-FFL E E I We suggest that auditory noise can regulate the temporal switching of
T4-FFL E I I bistable perception in neural circuit level.
T5-FFL I E E
T6-FFL I I E detector using random dot rotary motion stimuli. PLoS One 2009, 4(2):
T7-FFL I E I e4536.
T8-FFL I I I
E and I refer to excitatory and inhibitory neurons, respectively
Page 129 of 200
vector Y. We assumed a linear relation between input and output as
XW = Yand solved for the weights (the decoder parameters) using the
pseudo-inverse solution W = X Y. Then, given a 14-tuple vector
representation, z, of a discharge, the predicted object id is given
 
by ypred = arg min [0, 1, 2, 3, 4, 5] zT W , where 0 and 5
i=0..5
indicates a definite wrong prediction. For execution-only and
observation-only experiments, leave-one-out cross validation was applied
to obtain the success rates in decoding. For cross-decoding analysis, the
weight vector W obtained in one condition was used to predict the
object type in the other condition by using the data from that condition.
Acknowledgements: This work was supported by the grant
OBSERVENEMO within the framework of the bilateral S&T Cooperation
Program between the Republic of Turkey and the Hellenic Republic. Grant
no 113S391 funded by TUBITAK and grant 14UR OBSERVENEMO
co- Financed by the European Union and the Greek State, MCERA/GSRT.
References
1. Fadiga L, Fogassi L, Rizzolatti G: Action recognition in the premotor
cortex. Brain 1996, 119:593-609.
2. Papadourakis V, Raos V: Cue-dependent action-observation elicited
responses in the ventral premotor cortex (area F5) of the macaque
monkey. Soc Neurosci Abstr 2013, Program No. 263.08.
P191
Auditory noise influences human visual perception of ambiguous
information: multi-modal integration during bistable perception
Woochul Choi*, Se-Bum Paik
Department of Bio and Brain Engineering, KAIST, Daejeon 305-338, Republic
of Korea
E-mail: choiwc1128@kaist.ac.kr
BMC Neuroscience 2015, 16(Suppl 1):P191
When the sensory system receives an ambiguous signal, human perception
often switches spontaneously between two different interpretations. This
phenomenon is called bistable perception, and has been considered
important to understanding sensory system. In this study, we investigated
the intervals of spontaneous switching, defined as reversal time , to
examine the temporal dynamics of bistable perception. We also studied
the multi-modal feature of bistable switching by applying auditory noise
with the visual stimuli. Our hypothesis is that auditory noise would
significantly alter the reversal time of bistable visual perception. By
building a computational model, we could explain the influence of
auditory noise on the reversal time.
In the human psychophysical experiments, we first measured the reversal
time with visual stimulus only, using two types of bistable visual movies:
the racetrack [1] and the rotating 3D cylinder (Figure 1A). We observed
that the reversal times are widely varied across the subjects but fairly
consistent within the subject in both cases. Interestingly, we also found
that the reversal time for the racetrack and the rotating cylinder were
highly correlated (N = 9, R 2 = 0.84, Figure 1C). Next, we performed the
experiment with auditory noise and visual stimuli together, and found that
the reversal times are significantly altered. Importantly, when auditory
noise was given, we found a systematic change such that a fast switching
subjects (short ) tend to slow down the switching while slow switching
subjects (long ) tend to speed up, so that the difference of between the
two groups become insignificant (Figure 1D). Lastly, we designed a
double-well energy model with destabilization/restabilization processes [2],
and the model could well explain the observed phenomena.
perception in visual system and provides evidence of multi-modal
integration process in the sensory perception for ambiguous information.
References
1. Jain S: Performance characterization of Watson Ahumada motion
2. Kornmeier J, Bach M: Ambiguous figures - what happens in the brain
when perception changes but not the stimulus. Front Hum Neurosci 2012,
6:51.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P191) Effect of auditory noise in visual bistable perception. A. Two bistable visual movies. B. Example response and distribution.
C. Peak of distribution in no noise condition. D. Peak of distribution with auditory noise condition. Auditory noise significantly changed the reversal
time and the difference between the two groups has been decreased
P192
Local interaction in retinal ganglion cell mosaics can generate a
consistent spatial periodicity in cortical functional maps
Jaeson Jang*, Se-Bum Paik
Department of Bio and Brain Engineering, KAIST, Daejeon, 305-701, Republic
of Korea
E-mail: jaesonjang@kaist.ac.kr
BMC Neuroscience 2015, 16(Suppl 1):P192
Orientation map is one of the most studied functional maps in visual
cortex, but the developmental mechanism of its consistent spatial
periodicity is still elusive. Recently, a theoretical model suggested that a
moir interference pattern between ON and OFF retinal ganglion cell
(RGC) mosaics can develop a quasi-periodic orientation map, but it is
remained unclear how this can explain the constant periodicity of the
maps [1]. Here we suggest a developmental model that a simple local
interaction in RGCs can generate a consistent spatial periodicity of
orientation preference, by inducing (i) a hexagonal lattice structure in ON/
OFF RGC mosaics and (ii) a constant alignment angle between them.
First, we introduced a developmental model of a monotypic RGC mosaic to
show that a local repulsive interaction can generate a hexagonal structure
(Figure 1A). Previously, in the model study of the pairwise interaction point
process, it was suggested that a local interaction alone cannot develop a
long-range order in the mosaic structure [2]. We assumed a different type of
local repulsive interaction that the cell positions can be gradually shifted by a
Page 130 of 200
repulsion from the neighbor cells and confirmed that this model can develop
a long-range ordered structure that is well fitted to a hexagonal lattice.
Next, we assumed that there also exists a heterotypic repulsive interaction
between ON and OFF RGC mosaics and examined how this can affect the
alignment between the two mosaics (Figure 1B). When the inter-layer
distance between ON/OFF mosaics was varied within a proper interval, the
hexagonal structure was preserved in each mosaic, but the alignment angle
() between the two mosaics was restricted within a certain range of angles,
and this induced a constant spatial periodicity in the ON/OFF interference
pattern (Figure 1C). As observed in the moir interference, we confirmed a
consistent hexagonal periodicity in the cortical orientation map that are
simulated by statistical wiring model from the developed RGC mosaics
(Figure 1D,E) [3].
Conclusions: Our result suggests that a local repulsive interaction in RGC
mosaics can generate a hexagonal structure in ON/OFF RGC mosaics and a
restricted alignment between them. The interference between mosaics
induces a consistent spatial periodicity in cortical orientation map as
predicted by the moir interference pattern.
References
1. Paik SB, Ringach DL: Retinal origin of orientation maps in visual cortex.
Nat Neurosci 2011, 14:919-925.
2. Hore VR a, Troy JB, Eglen SJ: Parasol cell mosaics are unlikely to drive the
formation of structured orientation maps in primary visual cortex. Vis
Neurosci 2012, 29(6):283-299.
3. Ringach DL: Haphazard wiring of simple receptive fields and orientation
columns in visual cortex. J Neurophysiol 2004, 92(1):468-476.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P192) Simulation of RGC mosaics and orientation map. A. Development of monotypic mosaic and auto-correlation. B. Heterotypic
interaction depends on inter-layer distance. C. Moir interference pattern of ON/OFF RGC mosaics. D. Simulated orentation map. E. Hexagonal pattern in
auto-correlation of orientation map
P193
How bifurcations affect functional connectivity in finite-size neural
networks
Anna Cattani*, Diego Fasoli, Stefano Panzeri
Neural Computation Laboratory, Center for Neuroscience and Cognitive
Systems@UniTn, Istituto Italiano di Tecnologia, Rovereto, Italy
E-mail: anna.cattani@iit.it
BMC Neuroscience 2015, 16(Suppl 1):P193
Understanding how the functional connectivity of a neural network (i.e. the
statistical dependencies among different neurons) depends upon its
anatomical connectivity and how it is modulated by other network
parameters is a central topic in neuroscience [1]. However, only few
analytical investigations of functional connectivity in network models have
been reported [2]. A difficulty for such analytical studies is that they
require in principle characterizing the bifurcation points (i.e. the parameter
values at which the network undergoes a sudden change in dynamics) and
how functional connectivity changes close to such points. Given that
biological networks have a finite number of neurons, it is also important to
develop such analytical studies without relying only on large network
approximations that neglect finite-size effects. To overcome these
difficulties, here we developed a new approach to investigate the
dynamics of finite-size networks composed by interconnected inhibitory
and excitatory neurons. We applied this approach to a network of neurons
described by voltage-based differential equations, with an input to each
neuron provided by a deterministic current superimposed to a weak
normally distributed stochastic component.
To understand the interplay between the deterministic dynamics of the
network and the effects of random fluctuations in a finite-size network,
we proceeded in two steps.
First, we neglected the stochastic component of the input and we
performed a complete numerical and analytical bifurcation analysis while
varying the deterministic inputs and the strength of the connections. We
did this using the MatCont continuation package [3] and the analytical
expressions of the Jacobian matrix eigenvalues. We found that our finite-
size network displayed standard bifurcations such as saddle-node and
Hopf manifolds. Specifically, these manifolds describe the input currents
values for which equilibria collide and annihilate to each other, and for
which self-sustained oscillations arise, respectively. Moreover, strong
inhibition gave rise to additional branch point bifurcations: for highly
negative synaptic weights, the system exhibited special multiple solutions
Page 131 of 200
characterized by heterogeneous membrane potentials. These particular
solutions represent new finite-size effects not captured by large network
approximations.
Second, we inserted stochastic components into the input and studied the
behaviour of the network under the effect of noise. Specifically, we used a
new first-order perturbative method where the perturbative parameter is
the standard deviation of the noise in the input. This technique allowed the
analytical calculation of the cross-correlation between firing rates, and of the
Fisher information about the external input carried by the network. Our
analytical solutions showed that the neurons become independent when
the deterministic component of the input is strong, while they become
correlated for smaller current values lying on the saddle-node manifold. The
latter is due to a counter-intuitive finite-size effect [4] that can be fully
captured by our formalism. Furthermore, in our network strong correlations
led to large Fisher information values, thus enhancing the encoding
precision of the network with respect to the input variables. This also shows
that reducing correlations among neurons activity does not imply an
enhancement of the information encoding, as suggested in some studies [5].
Acknowledgements: This work was supported by the Autonomous
Province of Trento, Call Grandi Progetti 2012, project Characterizing
and improving brain mechanisms of attentionATTEND.
References
1. Buice MA, Chow CC: Dynamic Finite Size Effects in Spiking Neural
Networks. PLoS Computational Biology 2013, 9(1):e100287.
2. Pernice V, Staude B, Cardanobile S, Rotter S: Recurrent Interactions in Spiking
Networks with Arbitrary Topology. Physical Review E 2012, 85:031916.
3. Dhooge A, Govaerts W, Kuznetsov YA: Matcont: A matlab package for
numerical bifurcation analysis of ODEs. ACM Trans Math Softw 2003,
29(2):141-164.
4. Steyn-Ross ML, Steyn-Ross DA, Sleigh JW: Modelling general anaesthesia
as a first-order phase transition in the cortex. Progress in Biophysics &
Molecular Biology 2004, 85(2-3):369-385.
5. Ecker AS, Berens P, Keliris GA, Bethge M, Logothetis NK, Tolias AS: Decorrelated
Neuronal Firing in Cortical Microcircuits. Science 2010, 327(5965):584-587.
P194
Canonical correlations reveal co-variability between spike trains and
local field potentials in area MT)
Jacob Yates1, Evan Archer2, Alexander C Huk1, Il Memming Park3*
1
Center for Perceptual Systems, The University of Texas at Austin, Austin, TX
78712, USA; 2Department of Statistics and Grossman Center for the Statistics
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
of Mind, Columbia University, New York, NY 10027, USA; 3Department of
Neurobiology and Behavior, Stony Brook University, Stony Brook,
NY 11794, USA
E-mail: memming.park@stonybrook.edu
BMC Neuroscience 2015, 16(Suppl 1):P194
Patterns of neuronal correlations can provide important clues about the
structure of the underlying network and how it processes information.
Several recent studies have found that neural population activity across a
region can be explained in large part by a shared, low-dimensional signal
[1-5]. Population-wide correlation is likely to influence the local field
potential (LFP) - an epiphenomenon that reflects low-frequency, concerted
neural activity from anatomically connected circuits. Here, we show that
LFP and spike trains recorded simultaneously from the middle temporal
(MT) area of the awake macaque indeed share population-wide correlation.
We apply canonical correlation analysis (CCA) to 16 channels of LFP and 16
spike sorted neurons (from 12 channels) acquired at 50 ms temporal
resolution during inter-trial intervals (when the monkey was free to make
eye movements), as well as during performance of a perceptual decision-
making task (when the monkey maintained fixation and discriminated the
direction of visual motion). CCA finds instantaneous linear projections of
the LFP that maximize the correlation to corresponding projections of the
population spike trains.
Previous studies have suggested using population spike rate as a proxy
for the local network state [3,5]. Applied to our dataset, we obtain a
correlation coefficient of -12% between population spike rate and the
mean LFP during inter-trial interval segments. In contrast, we obtain pairs
of canonical variables with corresponding canonical correlations 29%,
26%, and 21%. We then applied the extracted projections to the task-
relevant motion stimulus integration window. We find that the correlation
of the projections is maintained for the 1st (31%) and 3rd (18%)
components, but drops significantly for the 2nd component (7%)
indicating a task-specific decoupling of LFP and spikes in a subspace
uncovered by CCA. Upon further analysis, each CCA projection showed a
distinct stimulus encoding pattern in spike rate and LFP. We hypothesize
that CCA projections reveal functional, virtual units of information
processing.
The LFP is an important source of information when neural activity is
correlated. It can indicate the strength of correlations, and the common
input giving rise to such correlations. Additionally, the LFP provides
increased statistical power to analyses, especially in areas where large-
scale recording is anatomically difficult. CCA is a simple technique that
can reveal low-dimensional structure in the data, uncovering components
which maximize covariability between LFP and spike trains within MT.
References
1. Ecker AS, Berens P, Cotton RJ, Subramaniyan M, Denfield GH, Cadwell CR,
et al: State dependence of noise correlations in macaque primary visual
cortex. Neuron 2014, 82(1):235-248.
2. Goris RLT, Movshon JA, Simoncelli EP: Partitioning neuronal variability. Nat
Neurosci 2014, 17(6):858-865.
3. Carandini M: Soloists and choristers in a cortical population. COSYNE
workshop on Scalable models for high-dimensional neural data 2014.
4. Luczak A, Bartho P, Harris KD: Gating of sensory input by spontaneous
cortical activity. Journal of Neuroscience 2013, 33(4):1684-1695.
5. Okun M, Yger P, Marguet SL, Gerard-Mercier F, Benucci A, Katzner S, et al:
Population rate dynamics and multineuron firing patterns in sensory
cortex. Journal of Neuroscience 2012, 32(48):17108-17119.
P195
Local structure supports learning of deterministic behavior in recurrent
neural networks
Jonathan Binas1,2*, Giacomo Indiveri1,2, Michael Pfeiffer1,2
1
Institute of Neuroinformatics, University of Zurich, Zurich, Switzerland; 2ETH
Zurich, Zurich, Switzerland
E-mail: jbinas@ini.ethz.ch
BMC Neuroscience 2015, 16(Suppl 1):P195
Many aspects of behavior, such as language, navigation, or logical
reasoning require strongly deterministic and sequential processing of
sensory and internal signals. This type of computation can be modeled
conveniently in the framework of finite automata.
Page 132 of 200
In this study, we present a recurrent neural network based on biologically
plausible circuit motifs, which is able to learn such deterministic behavior
from sensory input and reinforcement signals. We find that simple,
biologically plausible structural constraints lead to optimized solutions
and significantly improve the training process.
Previous work [1,2] has shown how arbitrary finite automata can be hand-
crafted in simple networks of neural populations by interconnecting multiple
Winner-Take-All units - small circuit motifs that match the properties of
cortical canonical microcircuits [3,4]. Figure 1 illustrates this transformation
from an automaton to neural network with populations of neurons encoding
either the state or potential state transitions. We extend that work by
introducing a reinforcement learning mechanism whose weight updates take
the form of reward-modulated Hebbian rule. This mechanism leads to
reconfiguration of the network connectivity in such a way that a desired
behavior is learned from sequences of inputs and reward signals.
As a key result of our study, we find that simple constraints on the
network topology, favoring local connectivity patterns, lead to dramatic
improvements both in training time and in the optimality of the found
solution, where the optimum is defined as the automaton with the
minimum number of states used to implement a given behavior. These
structural constraints correspond well to biological neural systems, where
short-range connections far outnumber long-range ones.
References
1. Rutishauser U, Douglas RJ: State-dependent computation using coupled
recurrent networks. Neural Comput 2009, 21(2):478-509.
2. Neftci E, Binas J, Rutishauser U, Chicca E, Indiveri G, Douglas JR:
Synthesizing cognition in neuromorphic electronic systems. Proc Natl
Acad Sci U S A 2013, 110(37):3468-3476.
3. Douglas RJ, Martin KAC: Neuronal circuits of the neocortex. Annu Rev
Neurosci 2004, 27(1):419-451.
4. Douglas RJ, Martin KAC: Recurrent neuronal circuits in the neocortex. Curr
Biol 2007, 17(13):496-500.
P196
Approximate nonlinear filtering with a recurrent neural network
Anna Kutschireiter1*, Simone C Surace1,2, Henning Sprekeler3,
Jean-Pascal Pfister1
1
Institute of Neuroinformatics, University of Zurich and ETH Zurich, 8057
Zurich, Switzerland; 2Department of Physiology, University of Bern, 3012
Bern, Switzerland; 3Bernstein Center for Computational Neuroscience,
Technical University Berlin, 10587 Berlin, Germany
E-mail: annak@ini.uzh.ch
BMC Neuroscience 2015, 16(Suppl 1):P196
One of the most fascinating properties of the brain is its ability to
continuously extract relevant features in a changing environment.
Realizing that sensory inputs are not perfectly reliable, this task becomes
even more challenging. This problem can be formalized as a filtering
problem where the aim is to infer the state of a dynamically changing
hidden variable given some noisy observation. A well-known solution to
this problem is the Kalman filter for linear hidden dynamics or the
extended Kalman filter for nonlinear dynamics. On the other hand, particle
filters offer a sampling-based approach to approximate the posterior
distribution. However, it remains unclear how these filtering algorithms
may be implemented in neural tissue. Here, we propose a neuronal
dynamics which approximates non-linear filtering.
Starting from the formal mathematical solution to the non-linear filter
problem, the Kushner equation [1], and assuming linear and noisy
observations we derive a stochastic rate-based network whose activity
samples the posterior dynamics. We found that taking samples following
these stochastic posterior dynamics is able to solve the inference task with a
performance comparable to that of standard particle filtering or (extended)
Kalman filtering. Indeed, for a linear hidden dynamics we exactly retrieve
the Kalman filter equations from our neural filter. In Figure 1 we show the
error of the filtered estimate as a function of the observation noise for two
different parameter choices in our filter equations.
Thus, the neuronal filter we propose provides an efficient way to infer the
state of temporally changing hidden variables. In addition, due to the
locality of the underlying mathematical model, the filter is made biologically
plausible from a neural-sampling perspective, hence providing a possible
framework for the neural sampling hypothesis [2].
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 133 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

Figure 1(abstract P195) Abstract graphical representation of an example finite automaton (A) and corresponding neural implementation (B).
The dotted boxes represent Winner-Take-All circuits in which only one population is active at a time. The populations labeled x represent the states while
the populations labeled y implement the conditional transitions between states. The colored connections are learned during training

Figure 1(abstract P196) Left: A sample trajectory of the real hidden state and its filtered estimate, showing the ability of the neural filter to
infer the hidden variable. Right: For a nonlinear hidden dynamics, the neuronal filter we propose achieves an estimation error which is comparable to
that of a particle filter or an extended Kalman filter (EKF). The worst neuronal filter corresponds to our filter with a suboptimal parameter choice

References
1. Kushner H: On the Differential Equations Satisfied by Conditional
Probability Densities of Markov Processes, with Applications. Journal of
the Society for Industrial & Applied Mathematics. Control 1962, 2(1).
2. Fiser J, Berkes P, Orbn G, Lengyel M: Statistically optimal perception and
learning: from behavior to neural representations. Trends in Cognitive
Sciences 2010, 14(3):119-130.
P197
The role of microcircuits in the pre-frontal cortex in detecting and
encoding temporally patterned information
Constantinos Melachrinos*, Athanasia Papoutsi, Panayiota Poirazi
IMBB, FORTH, Heraklion, 70013, Greece
E-mail: melachrinos@gmail.com
BMC Neuroscience 2015, 16(Suppl 1):P197
Working memory is the capacity of the brain to hold information temporarily
for immediate use. The neural correlate of this type of short-term memory is
persistent spiking activity (PA) of both excitatory and inhibitory neurons,
mainly in the pre-frontal cortex (PFC). Neurons embedded in PFC
microcircuits integrate widespread information from various brain regions.
PFC microcircuits of ~ten neurons are characterized by highly reciprocal
connections and non-linear integration in their dendrites. Understanding
how dendrites integrate information from multiple sources is crucial to
explain their functional role. Spatial and temporal integration of signals
occurs at dendrites before propagating to the soma, and plays a role in
coding of information [1,2]. The main question is whether and how
pyramidal neurons in such circuits can detect temporally structured
information in order to timely adjust behavior. Dendrites of cortical
pyramidal neurons have been shown to exhibit temporal sensitivity [1].
To address this question, we constructed a detailed PFC microcircuit,
implemented in the NEURON simulation environment. We used reconstructed
morphologies of layer 5 PFC pyramidal neurons, validated against
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
experimental findings. The microcircuit consisted of nine pyramidal neurons
and two interneurons, all interconnected [3]. To investigate temporal coding,
we delivered temporally structured input to four out of the nine pyramidal
neurons and assessed, given persistent activity emergence, a) the time-to-
first-spike (ttfs) of each pyramidal cell and b) the Inter-Spike Intervals (ISIs)
during PA.
We find that temporally patterned inputs (simulated as different temporal
orders of activated neurons) induce different responses, in both the ttfs
and ISI distributions. To investigate the mechanisms that mediate this
type of coding, we varied both the stimulus frequency and the pyramidal
neuron morphologies in the microcircuit. We found that lower stimulus
frequencies resulted in increased differences between various temporal
orders of activation. The same result was observed when using neurons
with morphologically complex dendritic trees, indicating that dendritic
morphology may play a key role in the ability of PFC microcircuits to
detect and encode temporally patterned inputs. Further, we investigated
whether this type of temporal coding exhibits specificity, i.e. whether the
PFC can consistently interpret similar temporally patterned inputs in a
spectrum of background activity environments.
Overall, this study seeks to understand at what level and how neuronal
circuits implement the timely firing of PFC pyramidal neurons. Improving
our understanding of temporal coding in complex areas like the PFC is
essential for disentangling how the brain detects, encodes and transmits
information. While dendrites of cortical neurons have been shown to
detect differences in the order of incoming signals, our study breaks new
ground in finding whether such a temporal code is preserved at the
microcircuit level. These findings can be tested experimentally to further
investigate the role of temporal coding in higher-level areas of the brain.
References
1. Branco T, Clark BA, Hausser M: Dendritic Discrimination of Temporal Input
Sequences in Cortical Neurons. Science 2010, 329(5999):1671-1675.
2. Poirazi P, Brannon T, Mel BW: Pyramidal Neuron as Two- Layer Neural
Network. Neuron 2003, 37(6):989-999.
3. Papoutsi A, Sidiropoulou K, Poirazi P: Dendritic nonlinearities reduce
network size requirements and mediate ON and OFF states of persistent
activity in a PFC microcircuit model. PLoS Computational Biology 2014,
10(7):e1003764.
P198
Multiplexed coding through synchronous and asynchronous spiking
Milad Lankarany1,2*, Steven A Prescott1,2
1
Neurosciences and Mental Health, The Hospital for Sick Children, Toronto,
ON, Canada; 2Department of Physiology and Institute of Biomaterials and
Biomedical Engineering, University of Toronto, Toronto, ON, Canada
E-mail: milad.lankarany@sickkids.ca
BMC Neuroscience 2015, 16(Suppl 1):P198
The prodigious capacity of our brain to process information relies on
efficient neural coding strategies. In engineered systems, bandwidth is
often increased through multiplexing - multiple signals are simultaneously,
yet independently, transmitted through a single communication channel.
We have proposed previously that neural systems might implement the
same sort of solution [1]. Here, we tested if/how multiplexed coding could
be achieved through combined rate and temporal coding. We
hypothesized that a set of neurons could independently encode two
signals by using asynchronous spike rate to encode one signal and
synchronous spike timing to encode the other.
To test our hypothesis, we built a feed-forward neural network comprising
Morris-Lecar (ML) model neurons. All neurons received a common input
constructed from two distinct signals, slow and fast, plus uncorrelated fast
noise. According to our hypothesis, slow and fast signals are
independently encoded by different types of spikes; in other words,
differentially correlated output spikes, namely, asynchronous (Async) and
synchronous (Sync), enable encoding of slow and fast signals, respectively.
To assess the feasibility of the multiplexed coding scheme, recorded spikes
were classified into two independent classes based on the peristimulus
time histogram (PSTH) calculated from the entire set of neurons. Spikes
whose instantaneous rates exceeded a threshold were designated Sync
and all others were designated Async. The spike triggered average (STA)
was calculated for slow and fast signals using Sync and Async spikes,
resulting in four different STAs. The Async-slow and Sync-fast STAs were
Page 134 of 200
clearly structured whereas the other two were not (Figure 1A). Using the
two spike types to calculate STAs from the single combined fast-slow input
gave two similarly structured STAs (meaning that one does not need a
priori knowledge of which signal evokes which spikes) whereas treating all
spikes as equivalent compromised STA measurement, especially for the
fast signal (meaning that one needs to subdivide spikes according to their
level of synchrony). Beyond simply measuring STAs, Figure 1B shows that
convolving the appropriately measured STA with the synchronous or
asynchronous spike train enables excellent reconstruction of the fast and
slow signal, respectively.
Our results demonstrate the feasibility of multiplexed coding using
synchronous and asynchronous spiking. Decoding the two signals
requires that spikes are distinguished by their cross-correlation; this is
difficult to do with experimental data in which only a subset of neurons
are recorded, but it is straightforward for the brain using downstream
neurons that operate as coincidence detectors or integrators that
respond preferentially depending on input correlation.
Acknowledgements: This work was supported by the National Sciences
and Engineering Research Council of Canada.
Reference
1. Ratt S, Hong S, De Schutter E, Prescott SA: Impact of neuronal properties
on network coding: roles of spike initiation dynamics and robust
synchrony transfer. Neuron 2013, 78(5):758-72.
P199
Partial information decomposition as a unified approach to the
characterization and design of neural goal functions
Michael Wibral1,2*, William A Phillips3, Joseph T Lizier4, Viola Priesemann5,6
1
MEG Unit, Brain Imaging Center, Goethe University, Frankfurt 60528,
Germany; 2Ernst Strngmann Institute for Neuroscience, Frankfurt 60528,
Germany; 3School of Natural Sciences, University of Stirling, Stirling FK9 4LA,
UK; 4School of Civil Engineering, The University of Sydney, Sydney, NSW
2006, Australia; 5Department of Nonlinear Dynamics, Max Planck Institute for
Dynamics and Self-Organization, 37077 Gttingen, Germany; 6Bernstein
Center for Computational Neuroscience, 37077 Gttingen, Germany
E-mail: Wibral@em.uni-frankfurt.de
BMC Neuroscience 2015, 16(Suppl 1):P199
In many neural systems anatomical motifs are found repeatedly in different
places. Despite this repetition these motifs often seem to serve a
perplexing variety of functions. A prime example is the canonical
microcircuit, which is repeated across multiple cortical areas, but supports
a variety of functions from sensory processing and memory to executive
functions and motor control. The multiplicity of functions served by a
single anatomical motif suggests a common, but more abstract,
information processing goal underlying all the different functions.
Identifying this goal from neural recordings is a key challenge in
understanding the general principles of neural information processing. The
apparent diversity of functions makes it clear that this common goal
cannot be described using function-specific language (e.g. edge filters),
but calls for an abstract framework. Here, information theory is the obvious
candidate. Notable past approaches using information theoretic
descriptions of neural goal functions suggested to maximize the mutual
information between input and output [1], maximize the coherent mutual
information that all the inputs share about the output [2], or, very
generally, to minimize the free energy [3]. To facilitate these efforts, and to
better dissect the implications of existing neural goal functions, we
suggest to build on a recent progress in information theory, termed partial
information decomposition (PID). PID allows to measure which of a set of
inputs contributes either uniquely, redundantly or synergistically to the
output of a (neural) processing unit [4-7], and which fraction of the
outputs entropy remains unexplained by the input set. We show how
these measures can be used to identify an information theoretic footprint
of a neural goal function. Most importantly, these measures can quantify
how much of the information is modified rather than merely relayed when
passing through the neural processor [8]. This shifts the focus from
information transmission to more complex processing and allows a much
better understanding of the (theoretical?) capabilities of a neuron or neural
circuit. Using this approach we show how to better understand existing
neural goal functions using PID measures and provide an information
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P198) Multiplexed coding (A) STAs calculated from slow or fast signals using asynchronous or syncrhonous spikes. (B)
Reconstructed inputs (colored) using two classes of spikes and corresponding STAs versus original stimulus (black)
theoretic framework for the design of novel goal functions for artificial
neural networks.
References
1. Linsker R: Self-organization in a perceptual network. Computer 1988,
21(3):105-117.
2. Kay JW, Phillips WA: Coherent Infomax as a computational goal for
neural systems. Bull Math Biol 2011, 73(2):344-372.
3. Friston K, Kilner J, Harrison L: A free energy principle for the brain.
J Physiol Paris 2006, 100(1-3):70-87.
4. Williams PL, Beer RD: Nonnegative Decomposition of Multivariate
Information. ArXiv10042515 Math-Ph Physicsphysics Q-Bio 2010.
5. Bertschinger N, Rauh J, Olbrich E, Jost J, Ay N: Quantifying Unique
Information. Entropy 2014, 16(4):2161-2183.
6. Griffith V, Koch C: Quantifying Synergistic Mutual Information. Guided Self-
Organization: Inception Springer Berlin Heidelberg: Prokopenko M 2014,
159-190, [Emergence, Complexity and Computation, vol. 9].
7. Wibral M, Lizier JT, Priesemann V: Bits from Brains for Biologically-Inspired
Computing. Frontiers in Robotics and AI 2015 in press.
8. Lizier JT, Flecker B, Williams PL: Towards a synergy-based approach to
measuring information modification. Artificial Life (ALIFE), 2013 IEEE
Symposium on. IEEE 2013, S43-S51.
P200
Cooperation/supervision of a habit by a cognitive strategy in a goal-
directed navigational paradigm
Souhel Hanoune*, Jean-Paul Banquet, Philippe Gaussier, Mathias Quoy
EIS Lab, University of Cergy-Pontoise, ENSEA - CNRS, Paris, France
E-mail: souheil.hanoune@ensea.fr
BMC Neuroscience 2015, 16(Suppl 1):P200
The Stimulus-Response (S-R) theory and Tolmans Cognitive Theory of
behavior control both issued from behaviorism in the early 20th century
still provide a relevant general framework to account for animal reward-
based adaptive behavior. In this paper, we propose a new paradigm for
representing and implementing both the cognitive strategy and the S-R
habit strategy within a unitary coding frame. Based on a parallel learning
of both strategies, the model explains how the fast learning cognitive
strategy can supervise and accelerate the slow learning S-R habit strategy;
and also how. In late learning stages, the habit strategy can overcome the
cognitive. This parallel representation is inspired by the cortico-basal
functional loops [1] and the cooperation between the cognitive associative
loop, including the dorso-medial striatum and the mPF; and the sensory-
motor loop, associated to the sensory motor cortex in relation with the
dorso-lateral striatum.
The implementation of S-R habit strategy is based on a neural modified
version of the classical Q-learning and is based on the model of [2],
emulating the functioning of the sensory-motor loop. The states of the
model are represented by hippocampal transitions, representing
associations between two consecutive place-cells during the exploration of
Page 135 of 200
the environment, learned in the CA1-CA3 regions of the hippocampus. The
cognitive strategy is based on a map representation of the environment
namely the cognitive map [3]. Based on the association between learned
transitions, the cognitive map allows the back-propagation of a reward
within a tree, allowing the selection of the shortest path to the goal. While
the cognitive map is quickly learned, the Q-values associated with the
Q-learning are slower to acquire. On the other hand, the Q-learning tends
to be more accurate than the cognitive map when fully learned.
The model exploits this speed difference in its parallel learning. The fast
acquisition of the cognitive map allows the robot to quickly choose correct
paths to the goal, and thus the time convergence of the Q-learning
algorithm is optimized. The cooperation is based on the biasing of the
selected transition by the cognitive map and the Q-learning in parallel (see
Figure 1). In its early learning stage, the Q-learning biasing is too weak,
and the cognitive map is dominant (Figure 2. VS Figure 2b), inducing the
supervision of the S-R habit by the cognitive strategy. In the later learning
stages, the Q-learning is stronger and more precise. Cooperation of the
cognitive strategy and S-R habit enables a faster S-R learning; as shown in
Figure 2. The lesion studies (Figure 2c, Figure 2d) show that the system
maintains a coherent behavior event after the lesion of either of the
structures supporting the two strategies. Also, the time responses highlight
the superiority of the habit strategy after over-training (Figure 2.c VS
Figure 2.d).
Acknowledgements: This work was supported by the ANR-NEUROBOT
project (ANR-BLAN-SIMI2-LS-100617-13-01).
References
1. DeLong MR, Strick PL: Parallel organization of functionally segregated
circuits linking basal ganglia and cortex. Annual Rev of Neurosci 1986,
9(1):357-381.
2. Hirel J, Gaussier P, Quoy M, Banquet J-P, Save E, Poucet B: The
Hippocampo-cortical Loop: Spatio-Temporal Learning and Goal-oriented
Planning in Navigation. Neural Networks 2013, 43:8-21.
3. OKeefe J, Nadel L: The hippocampus as a cognitive map. Oxford University
1978.
P201
A minimum-error, energy-constrained neural encoder predicts an
instantaneous spike-rate code
Erik C Johnson1,2,3*, Douglas L Jones1,2,3,4, Rama Ratnam2,3,4
1
Department of Electrical & Computer Engineering, University of Illinois,
Urbana, IL 61801, USA; 2Beckman Institute for Advanced Science and
Technology, University of Illinois, Urbana, IL, 61801, USA; 3Coordinated
Science Laboratory, University of Illinois, Urbana, IL, 61801, USA; 4Advanced
Digital Sciences Center, Illinois at Singapore Pte. Ltd., Singapore
E-mail: ejohns24@illinois.edu
BMC Neuroscience 2015, 16(Suppl 1):P201
An action potential (spike) is metabolically expensive to generate, and it
is likely that selective pressure has been exerted on the nervous system
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 136 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

Figure 1(abstract P200) Cooperative architecture between the cognitive and the S-R habit with a representation of the respective lesions

Figure 2(abstract P200) Statistical results for the sessions of the different groups representing the different situations where the S-R habits
trategy is alone, the cognitive strategy is alone, or when one of the two is lesioned after cooperative learning

to generate energy-efficient neural codes [1]. An additional constraint in
sensory systems is that the encoding should also represent stimuli with
minimal error. We postulate that this leads to a trade-off between energy
expenditure and encoding error, and propose that an optimal neural
code should minimize encoding error subject to a constraint on the
energy expended. A first approximation of energy expenditure by a
spiking neuron is E = b + kNs, where E: expended energy, b: baseline rate,
k: cost per spike, and Ns: the number of spikes fired. Figure 1A depicts
the encoding of an input signal s(t) as a spike-train. Given an energy
constraint E, the goal is to obtain the best possible reconstruction r(t) of s
(t), with a simple low-pass filter h(t) = A exp(-t/) (mimicking a post-
synaptic cell membrane) such that average squared reconstruction error e
(t) = s(t) - r(t) (Figure 1A) is minimized. Previously we showed that
minimal error encoding, subject to a spike-rate constraint, is equivalent to
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P201) A. Schematic of energy-constrained, minimum-error encoding model. B. A perturbation of the fishs electric field (top) and
the response of primary electrosensory neuron (below, experimental and optimal coder). Model produces r(t) in close match with s(t) and optimal spike-
times. C. Smoothed PSTH of the spike trains, and the predicted instantaneous rate function (blue, obscured)
a non-resetting dynamic threshold spike-firing model [2] with an dynamic
threshold h(t) and firing level g. Here, we show that this encoder can be
interpreted as an instantaneous rate encoder with rate i(t) = (s(t)/ + s(t))/
A. The function closely approximates the PSTH. We tested the
instantaneous rate coder by predicting: i) spike-times (Figure 1B, see
raster), and reconstruction (Figure 1B, top panel), and ii) the smoothed
PSTH (Figure 1C) for single-neuron data from p-type primary
electrosensory neurons of a weakly-electric fish (Apteronotus
leptorhynchus). For signals with little variability, this predicts a rate-code
of signal intensity. For signals with high variability, the spike-rate is driven
by the changes in the signal (i.e., its derivative). We conclude that
optimal encoder can optimally time spikes while maintaining high coding
fidelity, which can be interpreted as an instantaneous rate code.
Acknowledgements: This research was supported by NSF EFRI-BSBA-
0938007, NSF IGERT 0903622, the College of Engineering, UIUC, and a
research grant for the Human Sixth Sense Programme at the Advanced
Digital Sciences Center from Singapores Agency for Science, Technology
and Research (A*STAR). Electric fish data was collected in the laboratory
of Mark E. Nelson, UIUC, through NIH R01MH49242 and NSF IBN-0078206.
References
1. Laughlin SB: Energy as a constraint on the coding and processing of
sensory information. Curr Opin Neurobiol 2001, 11(4):475-480.
2. Kobayashi R, Tsubo T, Shinomoto S: Made-to-order spiking neuron model
equipped with a multi-timescale adaptive threshold. Front Comput
Neurosci 2009, 3:9.
P202
Regular and irregular stimuli result in changes in mice eye movement
and cerebellar nuclei neuron model behavior
Tiina Manninen1,2*, TD Barbara Nguyen-Vu2, Jennifer L Raymond2
1
Department of Signal Processing, Tampere University of Technology,
Tampere, Finland; 2Department of Neurobiology, Stanford School of
Medicine, Stanford, CA 94305, USA
E-mail: tiina.manninen@tut.fi
BMC Neuroscience 2015, 16(Suppl 1):P202
The cerebellum plays an important role in motor control. The cerebellar
key players include Purkinje cells (PCs), mossy fibers, climbing fibers, and
parallel fibers. Each PC receives inputs from many parallel fibers and from
a single climbing fiber. Cerebellar outputs originate from the deep
cerebellar and vestibular nuclei. Cerebellar nuclei (CN) neurons receive
inhibitory inputs from PCs and excitatory inputs from mossy fiber and
climbing fiber collaterals. In this work, we studied how regular and
irregular, as well as synchronous and asynchronous, PC firing frequencies
affect the eye movements in mice and CN neuron model behavior.
Floccular PCs on one side of the head were optogenetically stimulated
(see, e.g., [1]). We used both regular and irregular Poisson distributed
stimulus trains of 10-90 Hz. As the experimental output, we measured the
horizontal eye movements. We compared the effects of the parameters of
the stimulus train in the actual eye movements elicited with the effects of
the same parameters on the computational CN neuron model. The CN
Page 137 of 200
neuron model [2,3] includes 517 compartments and it receives an
inhibitory input from 450 PC synapses originating from 1-450 individual
PCs and an excitatory input from 150 mossy fiber synapses. The model has
eight different types of ion channels represented with Hodgkin-Huxley
type equations. We used both regular and Poisson distributed PC and
gamma distributed mossy fiber spike trains as inputs. We ran the model in
NEURON simulation environment [4] and did all the data analysis in
MATLAB. We varied the input irregularity, synchrony, mean firing rate
(20-120Hz), and PC to CN neuron convergence (1-450). As the model
output, we measured the cerebellar nuclei neuron firing rate and mean
GABA conductance. The experiments and the model simulations exhibited
similar behavior for some but not all of the stimulus frequencies. This
suggests a need for further experimental and simulation studies to more
fully understand how the parameters of PC firing rate influence
downstream oculomotor circuits.
Acknowledgements: This study was supported by the Academy of
Finland (application number 126556), US National Institutes of Health
(grant RO1 DC004154), as well as the Emil Aaltonen Foundation and Otto
A. Malm Foundation.
References
1. Nguyen-Vu TB, Kimpo RR, Rinaldi JM, Kohli A, Zeng H, Deisseroth K,
Raymond JL: Cerebellar Purkinje cell activity drives motor learning. Nat
Neurosci 2013, 16(12):1734-1736.
2. Luthman J, Hoebeek FE, Maex R, Davey N, Adams R, De Zeeuw CI,
Steuber V: STD-dependent and independent encoding of input
irregularity as spike rate in a computational model of a cerebellar
nucleus neuron. Cerebellum 2011, 10(4):667-682.
3. ModelDB, Accession: 144523. [https://senselab.med.yale.edu/ModelDB].
4. Carnevale NT, Hines ML: The NEURON Book Cambridge: Cambridge
University Press, 1st 2006.
P203
Contributions from active dendritic conductances to the Local Field
Potential
Torbjrn V Ness1*, Michiel WH Remme2, Gaute T Einevoll1,3
1
Dept. of Mathematical Sciences and Technology, Norwegian University of
Life Sciences, s, Norway; 2Institute for Theoretical Biology, Humboldt
University Berlin, Berlin, Germany; 3Dept. of Physics, University of Oslo, Oslo,
Norway
E-mail: torbjorn.ness@nmbu.no
BMC Neuroscience 2015, 16(Suppl 1):P203
The Local Field Potential (LFP) is typically assumed to mainly stem from
synaptic inputs and their subthreshold dendritic processing. The role of
active dendritic conductances in shaping the LFP has often been ignored,
even though several ion channels are active in the subthreshold voltage
regime. Here we use biophysical modeling to investigate the impact of
active conductances and their spatial distributions on the LFP in the
subthreshold voltage regime.
We used models of neocortical pyramidal neurons with realistic
morphologies and systematically varied the spatial distributions of active
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
conductances and synaptic inputs. We linearized the active conductances
yielding the so-called quasi-active description thereby reducing the
number of model parameters and highlighting that active currents can
be divided into two classes: regenerative and restorative. In terms of the
filtering properties of the cell membrane, regenerative conductances
amplify low frequencies of the synaptic input current, whereas restorative
conductances dampen them.
The low-frequency components of the synaptic input current give the
largest contributions to the LFP, due to their larger distance between
synaptic currents and associated return currents. Since active
conductances can enhance or diminish a neurons response to low
frequencies, they are also expected to affect the LFP.
We indeed found that the effect of active conductances on the single-
neuron LFP is strongest for the lowest frequencies (< 10-30 Hz). The
impact was most pronounced when 1) the synaptic input is asymmetric
(i.e., either basal or apical), 2) the active conductances are distributed
non-uniformly with the highest channel densities near the synaptic input,
and 3) when the LFP is measured at the opposite pole of the cell relative
to the synaptic input. Restorative conductances are typically expressed in
the LFP as resonance peaks. However, these are considerably smaller
than the corresponding peaks for the membrane potentials and
transmembrane currents.
We compared the findings from our quasi-active models with published,
experimentally constrained pyramidal cell models that display various
active conductances. We found that in particular the hyperpolarization-
activated inward current, IH, can have a sizable impact on the shape of
the LFP, since it is a restorative current, strong in the apical dendrite, and
active at subthreshold voltages.
We conclude that subthreshold contributions of active currents to the LFP
of detailed pyramidal cell models can be very well captured by quasi-
active models containing one or two voltage-dependent currents, and
that their contribution can be a major factor in shaping the LFP for the
lowest frequencies.
P204
Cross-cultural differences in visual attention: a computational
modelling study
Eirini Mavritsaki1,2*, Panagiotis Rentzelas1
1
Department of Psychology, Birmingham City University, Birmingham,
B422SU, UK; 2Department of Experimental Psychology, University of Oxford,
Oxford, OX1 3UD, UK
E-mail: eirini.mavritsaki@bcu.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P204
Literature in visual perception has identified that there are cross-cultural
differences in visual perception [1]. Research comparing members of
interdepended and collectivist East Asian cultures with independent and
individualist European American cultures into picture perception showed
that East Asians are more likely to attend the perceptual field as a whole
and to focus on context and Westerns to focus on the salient foreground
objects [1]. Research on cross-cultural differences has focused on
investigating cross-cultural differences related to bottom-up information.
Furthermore, research that experimentally manipulated the cultural norms
of individualism and collectivism groups managed to attenuate cultural-
specific preferences for social factors beneficial in human motivation [2].
Investigating the underlying mechanisms involved in these differences is
very important as it can affect everyday tasks, advertisement and many
other aspects of our everyday life.
Here we present the first steps of this work, investigating the underlying
processes in cross-cultural differences using computational modelling
studies. The computational model is based on the spiking Search over
Space and Time (sSoTS) model [3], that has been used to simulate Visual
Attention task. sSoTS has incorporated mechanisms that allows us to
investigate both bottom-up and top-down processes. We show that
sSoTS can successfully simulate cross-cultural differences in Visual
attention involving bottom-up tasks. Moreover, we expand the studies by
making predictions from the computational modelling studies for cross-
cultural differences and top-down tasks.
Acknowledgements: The authors would like to acknowledge the use of
the Advanced Research Computing (ARC) in carrying out this work.
Page 138 of 200
References
1. Miyamoto Y, Nisbett RE, Masuda T: Culture and the physical environment -
Holistic versus analytic perceptual affordances. Psychological Science 2006,
17(2):113-119.
2. Hagger MS, Rentzelas P, Chatzisarantis NLD: Effects of individualist and
collectivist group norms and choice on intrinsic motivation. Motivation
and Emotion 2014, 38(2):215-223.
3. Mavritsaki E, et al: Bridging the Gap Between Physiology and Behavior:
Evidence From the sSoTS Model of Human Visual Attention. Psychological
Review 2011, 118(1):3-41.
P205
Synaptic transmission of spike trains with arbitrary interspike intervals
Alex D Bird1,2,3*, Magnus JE Richardson1
1
Warwick Systems Biology Centre, University of Warwick, Coventry, UK;
2
School of Life Sciences, University of Warwick, Coventry, UK; 3Warwick
Systems Biology DTC, University of Warwick, Coventry, UK
E-mail: a.d.bird@warwick.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P205
Short-term synaptic depression, caused by depletion of releasable
neurotransmitter vesicles, modulates the strength of neuronal connections
in an activity-dependent manner [1,2]. Quantifying the statistics of this
form of synaptic transmission requires the development of stochastic
models linking probabilistic neurotransmitter release with the spike-train
statistics of the presynaptic population [3,4]. A common approach has
been to model the presynaptic spike train as either regular or a memory-
less Poisson process [5] - few analytical results are available that describe
the behaviour of a depressing synapse when the afferent spike train has
more complex, temporally correlated statistics.
Recently, we have derived a series of results that allow for the fraction of
occupied release sites and the neurotransmitter release probability to be
calculated for a presynaptic spike train with arbitrary interspike interval
(ISI) statistics. The results take a particularly compact form when the
presynaptic spike times are generated by a renewal process, i.e. when the
ISIs are independent. This encompasses a broad range of models that are
currently used for circuit and network analyses, including the class of
integrate-and-fire models. Our approach also allows for the postsynaptic
voltage mean and variance to be calculated, which in turn allows for an
approximation of the firing rate of a neuron driven by depressing synapses
from non-Poissonian presynaptic neurons (Figure 1).
These results will allow for the incorporation of more complex and
physiologically relevant firing patterns into future analytic studies of
neuronal circuits and networks.
References
1. Eccles JC, Katz B, Kuffler SW: Nature of the endplate potential in
curarized muscle. J Neurophysiol 1941, 4(5):362-387.
2. Tsodyks M, Pawelzik K, Markram H: Neural networks with dynamic
synapses. Neural Comput 1998, 10(4):821-835.
3. Fuhrmann G, Segev I, Markram H, Tsodyks M: Coding of temporal
information by activity-dependent synapses. J Neurophysiol 2002,
87(1):140-148.
Figure 1(abstract P205) Predictions of postsynaptic firing rates as a
function of the regularity of the renewal process governing
presynaptic firing. Presynaptic interspike intervals are independently
identically gamma distributed with shape parameter a (increasing a
gives more regular firing) and mean 1/Ra. Simulated rates are shown as
circles with error bars. Analytic predictions are shown as solid lines
based on exact forms for the postsynaptic voltage mean and variance
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 139 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

4. la Rocha de J, Parga N: Short-term synaptic depression causes a non-

monotonic response to correlated stimuli. J Neurosci 2005, 25(37):8416-8431.
5. Rosenbaum R, Rubin J, Doiron B: Short Term Synaptic Depression
Imposes a Frequency Dependent Filter on Synaptic Information Transfer.
PLoS Comput Biol 2012, 8(6):e1002557.

P206
Coarse-grained description of the spatio-temporal dynamics of network
activity from experimentally verified single-neuron models and connectivity
Francesco Fermani*, Magnus JE Richardson
Warwick Systems Biology Centre, University of Warwick, Coventry, UK
E-mail: f.fermani@warwick.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P206

We combine experimentally constrained models of neocortical neuron

voltage dynamics [1,2] and network connectivity [3] to derive a set of
equations that describe the activity at a tissue scale. The resulting equations
represent a neuronal field theory in which emergent properties at a coarse-
grained level can be causally linked to the physiology of cellular and sub-
cellular components. The description is mathematically tractable and can be
elaborated to include further biophysical details such as multiple neuronal
populations to capture the structure of the component microcircuits,
synaptic dynamics and filtering as well as distance-dependent delays in
signal propagation. For spatially homogeneous afferent drive the steady-
state firing rate can be straightforwardly calculated together with the
network response to weak spatio-temporal modulation of the afferent drive
via a perturbative approach. For the spatially heterogeneous non-linear
regime, which the network is pushed into under stronger drive, we
construct an iterative numerical scheme that rapidly converges to the
network firing rate. The utility of the approach is illustrated using examples
from the experimental literature.
References
1. Fourcaud-Trocm N, Hansel C, van Vresswijk C, Brunel N: How Spike
Generation Mechanisms Determine the Neuronal Response to
Fluctuating Inputs. J Neurosci 2003, 23(37):11628.
2. Badel L, Lefort S, Brette R, Petersen CCH, Gerstner W and Richardson MJE:
Dynamic I-V curves are reliable predictors of naturalistic pyramidal-
neuron voltage traces. J Neurophys 2008, 99(2):656.
3. Perin R, Berger T K, Markram H: A synaptic organizing principle for cortical
neuronal groups. PNAS 2011, 108(13):12.
P207
On-line identification of the end of motor imageries based on the
alpha rebound detection
Cecilia Lindig-Len1,2,3*, Laurent Bougrain1,2,3, Sbastien Rimbert1,2,3
1
Inria, Villers- ls -Nancy, F-54600, France; 2Universit de Lorraine, LORIA,
UMR 7503, Vanduvre-ls-Nancy, F-54500, France; 3CNRS, LORIA, UMR 7503,
Vanduvre-ls-Nancy, F-54500, France
E-mail: cecilia.lindig-leon@inria.fr
BMC Neuroscience 2015, 16(Suppl 1):P207
Limb movement execution and imagination elicit in a mutually exclusive
manner sensorimotor rhythms that can be detected in electroencephalographic
(EEG) recordings; in particular over the primary motor cortex, where an
oscillatory modulation has been observed prior, during and following the
execution of voluntary movement, passive movement, imagined movement,
and even tactile stimulation [1,2]. The modulation following the movement
termination consists of an event-related synchronization (ERS) that increases the
oscillatory power for a few hundred milliseconds [3]. Since it is known to be
specific of the beta band 13-25 Hz, it is denoted as post-movement beta
rebound, although in recent studies it has been shown that this phenomenon is
enhanced when analyzed in the alpha range 8-13 Hz (see Figure 1) [1,4].
The characteristics of this post-movement rebound, as it will be shown in
the present study, are preserved independently of the involved limb during
the motor execution. From database 2a of the BCI competition IV [4], an on-
line method for identifying the end of motor imageries on a single trial
detection is presented. By using an overlapped sliding window over each
trial from four different motor imageries (left hand, right hand, feet and
tongue), two contrasting classes are generated according to the occurring
condition (i.e., segments with rebound and segments without it) to generate
Figure 1(abstract P207) Averaged post-movement rebound for
subject 7 in electrode C4 across the A. alpha and B. beta bands. The
yellow box indicates the duration of the motor imageries
a classification model based on a linear discriminant analysis. Results show
that the classification performance is 5% superior over the alpha band than
the beta band for almost all subjects, and that the rebound detection is
independent from the limb used in the motor imagery.
Conclusions: On-line detection of the end of motor imageries of various
body parts is feasible by detecting the post-movement alpha rebound. The
accuracy reached by the proposed method within the alpha band across
all subjects is 79.17% with a sensitivity value of 0.81 and specificity of 0.71.
This method improves the detection of the end of motor imageries by
considering the alpha post-movement rebound, which is of interest for the
design of self-paced brain-computer interfaces.
References
1. Avanzini P, Fabbri-Destro M, Dalla Volta R, Daprati E, Rizzolatti G,
Cantalupo G: The dynamics of sensorimotor cortical oscillations during
the observation of hand movements: an EEG study. PLoS One 2012, 7(5):
e37534.
2. Pfurtscheller G, Stanck A: Post-movement beta synchronization and
desynchronization. A correlate of an idling motor area?
Electroencephalogr Clin Neurophysiol 1996, 98(4):281-93.
3. Lindig-Len C, Bougrain L, Rimbert S: Alpha rebound improves on-line
detection of the end of motor imageries. International IEEE/EMBS
Conference on Neural Engineering 2015 [https://hal.inria.fr/hal-01092284].
4. Tangermann M, Mller KR, Aertsen A, Birbaumer N, Braun C, Brunner C,
et al: Review of the BCI competition iv. Front Neurosci 2012, 6:55.
P208
Optimal signal detection with neuronal diversity: balancing the gullible
and the prudent neurons
Leonardo L Gollo1*, Mauro Copelli2, James A Roberts1
1
Systems Neuroscience Group, QIMR Berghofer, Brisbane, Queensland,
Australia; 2Departmento de Fsica, Universidade Federal de Pernambuco,
Recife, Pernambuco, Brazil
E-mail: leonardo.l.gollo@gmail.com
BMC Neuroscience 2015, 16(Suppl 1):P208
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P208) Top: Illustrative random networks with neuronal diversity in the threshold parameter . Bottom: Maximal dynamic range
max (as defined in ref. [6]) reached by networks with binomial (left), uniform (center), and gamma (right) distributions (bottom). Networks with binomial
distribution have a proportion of integrator neurons with =2, whereas the remainders are non-integrator neurons (=1). The threshold In the uniform
distribution varies from 1 to max. Network size is 5000 neurons
Network connectivity have been shown to play an important role in
shaping the neuronal dynamics [1-5]. A complementary remarkable feature
of neuronal systems is the large degree of morphological and functional
diversity. Despite some recent efforts in understanding the role of
neuronal diversity embedded in a network [6-9], the benefits of cellular
variability to distinguish input varying over orders of magnitude remain
elusive. We utilize a simple quiescent-active-refractory-quiescent model,
which is amenable to mathematical analysis [10], interacting in a (non-
structured) random network with diversity in the parameter that controls
the propensity of the neurons to fire in response to input from their
neighbors. We consider a simple binomial distribution, a uniform
distribution, and a more realistic gamma distribution. As depicted in Figure
1, we show that the capability of the network to distinguish the amount of
external input can be improved by two orders of magnitude (20 dB) in the
presence of diversity. We explain how diversity enhances the network
capabilities, and identify the cases in which one specialized sub-population
outperforms the rest of the network and the cases in which the average
network outperforms any sub-population. Finally, we show the robustness
of our results in a balanced cortical network of excitatory and inhibitory
neurons.
References
1. Gollo LL, Zalesky A, Hutchison RM, van den Heuvel M, Breakspear M:
Dwelling quietly in the rich club: brain network determinants of slow
cortical fluctuations. Phil Trans R Soc B 2015, 370(1668).
2. Matias FS, Gollo LL, Carelli PV, Bressler SL, Copelli M, Mirasso CR: Modeling
positive Granger causality and negative phase lag between cortical
areas. NeuroImage 2014, 99:411-418.
3. Gollo LL, Mirasso C, Sporns O, Breakspear M: Mechanisms of zero-lag
synchronization in cortical motifs. PLoS Comput Biol 2014, 10(4):e1003548.
4. Gollo LL, Breakspear M: The frustrated brain: from dynamics on motifs to
communities and networks. Phil Trans R Soc B 2014, 369(1653):20130532.
5. Moretti P, Muoz MA: Griffiths phases and the stretching of criticality in
brain networks. Nat Commun 2013, 4.
Page 140 of 200
6. Gollo LL, Mirasso C, Eguluz VM: Signal integration enhances the dynamic
range in neuronal systems. Phys Rev E 2012, 85(4):040902.
7. Mejias JF, Longtin A: Optimal heterogeneity for coding in spiking neural
networks. Phys Rev Lett 2012, 108(22):228102.
8. Vladimirski BB, Tabak J, ODonovan MJ, Rinzel J: Episodic activity in a
heterogeneous excitatory network, from spiking neurons to mean field.
J Comput Neurosci 2008, 25(1):39-63.
9. Tessone CJ, Mirasso CR, Toral R, Gunton JD: Diversity-induced resonance.
Phys Rev Lett 2006, 97(19):194101.
10. Gollo LL, Kinouchi O, Copelli M: Statistical physics approach to dendritic
computation: The excitable-wave mean-field approximation. Phys Rev E
2012, 85(1):011911.
P209
A cortical multi-layered model and the properties of its internally-
generated activity
Rodrigo FO Pena*, Renan O Shimoura, Antonio C Roque
Departamento de Fsica, FFCLRP, Universidade de So Paulo, Ribeiro Preto,
SP, Brazil
E-mail: rfdop20@gmail.com
BMC Neuroscience 2015, 16(Suppl 1):P209
The cerebral cortex displays a rich repertoire of internally-generated
dynamic states even in the absence of external stimuli [1]. Most theoretical
studies of cortical activity are based on networks of randomly connected
units [2] or with architectures artificially built from random networks [3]. In
spite of the usefulness of these models, it is also important to have
computational models that try to accurately represent cortical network
architecture. Recently, Potjans and Diesmann [4] presented a network
model of the local cortical microcircuit based on extensive experimental
data on the intrinsic circuitry of striate cortex. The model is a full-scale
representation of the cortical network under a surface area of 1 mm2 of
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
striate cortex (~80,000 neurons) and contains two cell types (excitatory and
inhibitory) distributed over four layers: L2/3, L4, L5, and L6. The cells are
modeled as current-based leaky integrate-and-fire neurons with
exponential synaptic currents. In this work, we used the connectivity map
of the Potjans and Diesmann model [4] to construct a cortical model with
4,000 neurons (i.e. with the number of cells reduced by a factor of 20 in
comparison with the Potjans and Diesmann model). Cells were described
by the Izhikevich model [5] with parameters adjusted so that excitatory
neurons were of the regular spiking (RS) type, 50% of the inhibitory
neurons were of the fast spiking (FS) type and the other 50% of the
inhibitory neurons were of the low-threshold spiking (LTS) type. Synapses
were modeled as conductance-based with exponentially decaying
conductances (we used the same synaptic parameters as in [3]).
Instantaneous excitatory/inhibitory synaptic increments were denoted by
gex/gin. Brief (10 ms) but strong direct current pulses were applied to 15%
of L4 excitatory cells to stimulate the network and, after stimulus removal,
we kept the simulation running until Tsim = 3000 ms. We performed this
experiment for 10 different initial conditions to randomize the construction
of the network as well as for at least 100 different combinations of gex/gin
in the range gex = [0, 0.1], gin = 0[1]. The measurements taken were the
lifetime of network activity, the activity of the network and the coefficients
of variation of the interspike intervals of network neurons. In addition, we
performed the same experiments with L4 isolated and in all possible
combinations (in pairs or triplets) with the other layers, and with 100% of
inhibitory cells of the FS type.
The major results of our simulations are: (1) For networks made of RS and
FS cell types, long-lived network activity was observed for combinations
of gex/gin in the region of highest values of both of them. These states
displayed irregular neuronal firing. For other combinations of gex/gin the
network activity decayed rapidly after a short transient; (ii) Introduction of
LTS neurons increased the region of gex/gin combinations that generated
long-lived activity and reduced the average network firing rate; (iii)
Different combinations of layers favored more or less the occurrence of
long-lived activity. L4 alone, L4-L5 and L4-L5-L6 could not sustain long-
lived activity while L4-L6 and L23-L4-L6 displayed long-lived activity for
larger regions of gex/gin combinations.
Acknowledgements: Work funded by FAPESP-DFG (IRTG 1740/TRP 2011/
50151-0) and FAPESP/CEPID/Neuromat (grant 2013/07699-0). RFOP is
supported by a FAPESP PhD scholarship (grant 2013/25667-8), ROS is
supported by a CAPES MSc scholarship, and ACR is supported by a CNPq
research grant (PQ 306251/2014-0).
References
1. Vogels TP, Rajan K, Abbott LF: Neural network dynamics. Annu Rev
Neurosci 2005, 28:357-376.
2. Kriener B, Enger H, Tetzlaff T, Plesser HE, Gewaltig M, Einevoll GT: Dynamics
of self-sustained asynchronous-irregular activity in random networks of
spiking neurons with strong synapses. Front Comput Neurosci 2014, 8:136.
3. Tomov P, Pena RFO, Zaks MA, Roque AC: Sustained oscillations, irregular
firing, and chaotic dynamics in hierarchical modular networks with
mixtures of electrophysiological cell types. Front Comput Neurosci 2014,
8:103.
4. Potjans TC, Diesmann M: The cell-type specific cortical microcircuit:
relating structure and activity in a full-scale spiking network model.
Cereb Cortex 2014, 24(3):785-806.
5. Izhikevich EM: Simple model of spiking neurons. IEEE Trans Neural Netw
2003, 14(6):1569-1572.
P210
Effect of synaptic plasticity on functional connectivity and global
activity of a neocortical network model
Renan O Shimoura*, Rodrigo FO Pena, Antonio C Roque
Departamento de Fsica, FFCLRP, Universidade de So Paulo, Ribeiro Preto,
SP, Brazil
E-mail: renanshimoura@usp.br
BMC Neuroscience 2015, 16(Suppl 1):P210
Neocortex plays key role in diverse brain functions. Understanding this
role involves the study of collective neural activity patterns under
different situations, and how these patterns relate to the structural and
functional organization of neocortex. Here we study the effect of synaptic
plasticity on neural spiking activity patterns in a neocortical network
Page 141 of 200
model. We measure changes in neural spiking patterns due to changes in
the strengths of the synapses connecting neurons and relate them to
changes in the functional connectivity of the network as disclosed by
graph-theoretic measures.
Our neocortical network model was composed of excitatory and inhibitory
neurons in the proportion of four excitatory cells for each inhibitory cell.
Neurons were described by the Izhikevich model [1]. The parameters of
the model were adjusted so that excitatory neurons were of the regular
spiking (RS) type and inhibitory neurons were all of either the fast spiking
(FS) or the low-threshold spiking (LTS) type. Synapses were modeled as
event-based, and two types of synaptic dynamics were considered: one
without synaptic plasticity in which the synaptic weight received a fixed
increment after the pre-synaptic event and decayed exponentially after
that, and one with synaptic plasticity in which the synapse obeyed an
asymmetric spike-timing dependent plasticity (STDP) rule described by [2].
Neurons were organized into four layers (2/3, 4, 5 and 6) with layer- and
cell-specific statistical connectivity rules based on [3]. The total number of
neurons in the model was about 4,000. Two experiments were done: one
with all synapses described by the model without synaptic plasticity, and
the other with synapses between excitatory neurons described by the
STDP rule while the remaining synapses were described by the model
without synaptic plasticity. In both cases, the model was stimulated by a
current injection of random amplitude applied to neurons of layer 4 (L4),
which is the main input layer of the cortex. The spiking activity of the
network was evaluated by measures extracted from the raster plot of the
spikes produced by the neurons, e.g. layer-specific and network mean and
time-dependent firing rates. The structural and functional connectivities of
the network were represented by the respective structural and functional
adjacency matrices. The functional adjacency matrix was constructed by
taking in consideration neuron pairs with strength of their synaptic
coupling above a specific threshold. The topology of the adjacency
matrices was characterized by graph-theoretic measures, e.g. clustering
coefficient.
We determined a set of parameters for which the spiking activity
generated in L4 by the external input propagated to the entire network.
This network-wide activity was oscillatory, and we found that its mean
frequency was higher for the network version with synaptic plasticity
than for the version without synaptic plasticity. We also found that the
formation of clusters of synchronous neural activity was facilitated in the
case with LTS cells as inhibitory neurons. Our results suggest that
synaptic plasticity may induce changes in the functional connectivity of
the neocortical network with impact on its global activity.
Acknowledgements: Work funded by FAPESP-DFG (IRTG 1740/TRP 2011/
50151-0) and FAPESP/CEPID/Neuromat (grant 2013/07699-0). ROS is
supported by a CAPES MSc scholarship, RFOP is supported by a FAPESP
PhD scholarship (grant 2013/25667-8), and ACR is supported by a CNPq
research grant (PQ 306251/2014-0).
References
1. Izhikevich EM: Simple model of spiking neurons. IEEE Trans Neural Netw
2003, 14(6):1569-1572.
2. Song S, Miller KD, Abbott LF: Competitive Hebbian learning through
spike-timing dependent synaptic plasticity. Nature Neurosci 2000,
3:919-926.
3. Potjans TC, Diesmann M: The cell-type specific cortical microcircuit:
relating structure and activity in a full-scale spiking network model.
Cereb Cortex 2014, 24(3):785-806.
P211
Dynamics of competition between coupled spiking networks in the
balanced state
Fereshteh Lagzi1,2*, Stefan Rotter1,2
1
Bernstein Center Freiburg, Freiburg, Germany; 2Faculty of Biology, University
of Freiburg, Freiburg, Germany
E-mail: fereshteh.lagzi@bcf.uni-freiburg.de
BMC Neuroscience 2015, 16(Suppl 1):P211
The nonlinear mean-field dynamics of spiking neuronal networks with
non-homogeneous connectivity is studied numerically and analyzed
mathematically. The network under study is comprised of three
subnetworks of either excitatory or inhibitory leaky integrate-and-fire
neurons, two of which are of the same type. The excitatory and inhibitory
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
weights are arranged to establish and maintain a balance between
excitation and inhibition for a constant external drive. Each subnetwork
has random connectivity with fixed in-degree and fixed out-degree for all
neurons belonging to a particular population (configuration model).
Neurons are also randomly connected with the same probability across
different subnetworks; however, depending on their identity, the
connection weight is scaled by a common factor. We observed that for a
certain regime of ratios of the within versus between connection
weights (bifurcation parameter), network activation spontaneously
switches between the two subnetworks of the same identity. In the
mean-field model, this phenomenon is explained by a set of coupled
stochastic differential equations of Lotka-Volterra type [1,2] that establish
competition between the subnetworks. The model shows the same
dynamical behavior as observed in simulations of large spiking networks.
The deterministic phase portrait is characterized by two attractors and a
saddle node, its stochastic component is essentially given by the
multiplicative inherent noise of the system [3]. We observe that the life
time distribution of the active states is exponential, therefore, and it
appears that noise fluctuations kicks the system from one attractor to the
other. The same model for a larger number of populations suggests a
general approach to study the dynamics of interacting populations of
spiking networks.
Acknowledgements: This work was supported by the German Ministry of
Education and Research (BFNT Freiburg*Tbingen, grant 01GQ0830) and
the German Research Foundation (DFG, grant EXC 1086). This work was
performed on the computational resource bwUniCluster funded by the
Ministry of Science, Research and Arts and the Universities of the State of
Baden-Wrttemberg, Germany, within the framework program bwHPC.
References
1. Lotka AJ: Analytical note on certain rhytmic relations in organic systems.
PNAS 1920, 6(7):410-415.
2. Volterra V: Variations and Fluctuations of the Number of Individuals in Animal
Species living together McGrawHill: Chapman, R.N 1931.
3. Lagzi F, Rotter S: A Markov model for the temporal dynamics of
balanced random networks of finite size. Frontiers in Computational
Neuroscience 2014, 8:1-23.
P212
The formation of habits in the neocortex under the implicit supervision
of the basal ganglia
Meropi Topalidou1,2,3, Daisuke Kase2,4, Thomas Boraud2, Nicolas P Rougier1,2,3*
1
INRIA Bordeaux Sud-Ouest, Bordeaux, France; 2Universit de Bordeaux, CNRS
UMR 5293, IMN, France; 3LaBRI, Universit de Bordeaux, IPB, CNRS, UMR
5800, Talence, France; 4Laboratoire Franco-Isralien de Neurosciences, CNRS
Bordeaux, Talence, Bordeaux, France
E-mail: Nicolas.Rougier@inria.fr
BMC Neuroscience 2015, 16(Suppl 1):P212
Figure 1(abstract P212) Left. In habitual condition (HC), performances are optimal, with or without GPi. In novel condition (NC), only the intact
model (with GPi) is able to learn the new stimuli while lesioned model performances stay at the level f chance. Right. Analysis of the data shows
that reaction time is higher in normal condition as compared to habitual condition with active and inactive GPi. The later increases significantly the
reaction time in both conditions. All data have been averaged over 250 simulations
Page 142 of 200
If basal ganglia are widely accepted to participate in the high-level
cognitive function of decision-making, their role is less clear regarding
the formation of habits [1,2]. One of the biggest problem is to
understand how goal-directed actions are transformed into habitual
responses, or, said differently, how an animal can shift from an action-
outcome (A-O) system to a stimulus-response (S-R) one, while
maintaining a consistent behavior. We introduce a computational model
(basal ganglia, thalamus and cortex) that can solve a simple two arm-
bandit task using reinforcement learning and explicit valuation of the
outcome [3]. Hebbian learning has been implemented at the cortical level
such that the model learns each time a move is issued, rewarded or not.
Then, by inhibiting the output nuclei of the model (GPi), we show how
learning has been transferred from the basal ganglia to the cortex, simply
as a consequence of the statistics of the different choices (see Figure 1).
Because best (in the sense of most rewarded) actions are chosen more
often, this directly impacts the amount of Hebbian learning and lead to
the formation of habits within the cortex. These results have been
confirmed in monkeys doing the exact same task where the BG has been
inactivated using muscimol. This tends to show that the basal ganglia
implicitly teach the cortex in order for it to learn the values of new
options. In the end, the cortex is able to solve the task perfectly, even if
it exhibits slower reaction times.
References
1. Yin HH, Knowlton BJ: The role of the basal ganglia in habit formation.
Nature Reviews Neuroscience 2006, 7(6):464-476.
2. Seger CA, Spiering BJ: A critical review of habit learning and the basal
ganglia. Frontiers in Systems Neuroscience 2011, 5:66.
3. Guthrie M, Leblois A, Garenne A, Boraud T: Interaction between cognitive
and motor cortico-basal ganglia loops during decision making: a
computational study. Journal of Neurophysiology 2013, 109(12):3025-3040.
P213
A realistic model of pitch explains the N100m morphology evoked by
dyads
Alejandro Tabas1*, Emili Balaguer-Ballester1,2, Andr Rupp3
1
Faculty of Science and Technology, Bournemouth University, Bournemouth,
UK; 2Bernstein Center for Computational Neuroscience, Heidelberg-
Mannheim, Baden-Wrttemberg, Germany; 3Heidelberg University, Baden-
Wrttemberg, Germany
E-mail: atabas@bournemouth.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P213
Pitch is a fundamental attribute of auditory sensation underlying the
perception of complex sounds. The N100m is a well-known transient
neuromagnetic response of the evoked fields observed in magneto-
encephalographic recordings, sensitive to fundamental properties of
auditory stimuli such as pitch or timbre [1]. Despite considerable empirical
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
evidence of the association between pitch perception and the N100m
deflection in antero-lateral Heschls gyrus (primary auditory cortex) [1], the
neurophysiological mechanisms of pitch processing in cortex are still
poorly understood. In this study we propose an innovative approach for
understanding the pitch processing-N100m association by combining a
biophysical model of the peripheral auditory system with a network of
neural ensembles using realistic neural and synaptic parameters. The
model is able to reproduce the morphology of the N100m component of
auditory evoked fields in humans.
The model consists of three stages. First, a biophysically realistic model of
the auditory periphery transforms the input stimuli into temporal patterns
of auditory nerve activity [2]. This step is followed by an autocorrelation
process that generates a spectro-temporal representation of the
peripheral activity. These two stages are common in the literature and
are supported by perceptual and neurophysiological data in sub-cortical
areas [2]. The third stage processes patterns of brainstem activity by
means of an ensemble model of 100 cortical populations parametrised by
preferred frequency values [3]. The ensemble model stems from a mean
field approximation of a network of spiking neurons, further simplified
such that NMDA receptors dominate the model dynamics [3]; which
yields to a network architecture with recurrent self-excitations and
effective inhibitory currents between ensembles [3].
Electrophysiological predictions of the N100m morphology were derived from
the aggregated dynamics of the gating variables of the cortical populations
[4]; whilst the predicted pitch was encoded in a preferred frequency
characterising a neural ensemble [2]. Biophysical parameters of the model
were chosen according to neurophysiological data [4]. Mutual inhibition
strengths, i.e. the effective connectivities between neural ensembles, were
fitted and validated using pure and harmonic complex tones. The remaining
parameters were tuned such that the model prediction for a unison dyad
(two simultaneous harmonic complex tones) matched the evoked N100m
morphology (10ms before and 70ms after the deflection). Model predictions
were then evaluated using two other different dyads (see Figure 1).
The model successfully predicts the N100m morphology associated to
such stimuli, for the first time to our knowledge. Furthermore, ensemble
connectivities naturally reveal a harmonic structure critical for the cortical
processing of pitch. Interestingly, the model dynamics unveils that that
the N100m deflection is the result of the succession of a large increase in
the input current of the neural populations followed by a selective
inhibitory process. Thus, our results suggest that the model can
potentially explain the biophysical mechanisms underlying a range of
neurophysiological data associated to pitch perception.
References
1. Ritter S, Dosch HG, Specht HJ, Rupp A: Neuromagnetic responses reflect
the temporal pitch change of regular interval sounds. NeuroImage 2005,
27(3):533-543.
2. Balaguer-Ballester E, Clark N: Understanding pitch perception as a
hierarchical process with top-down modulation. PLoS Computational
Biology 2009, 5(3):e1000301.
3. Wong KF, Wang XJ: A recurrent network mechanism of time integration
in perceptual decisions. The Journal of Neuroscience 2006, 26(4):1314-1328.
4. Deco G, Ponce-Alvarez A, Mantini D, Romani GL, Hagmann P, Corbetta M:
Resting-state functional connectivity emerges from structurally and
Figure 1(abstract P213) Model predictions and recorded evoked fields for an unison dyad (A), a trinone (B) and a perfect fifth (C)
Page 143 of 200
dynamically shaped slow linear fluctuations. The Journal of Neuroscience
2013, 33(27):11239-11252.
P214
The delayed response network: towards a single layer universal neural
network approximator and delay-based learning
Dinov Martin1*, Elias Rut2
1
Computational, Cognitive and Clinical Neuroimaging Laboratory, Imperial
College London, London, UK; 2Vienna University of Technology, Vienna,
Austria
E-mail: m.dinov13@imperial.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P214
We were interested in exploring the use of timing or delays in learning in
neural networks. The work was developed by wondering about the
effects of glia on network learning in the human brain. To investigate this
idea, we created a Multilayer Perceptron (MLP)-based [1] feedforward
model that included a signal propagation time (delay) parameter, which
allowed for bit patterns to be learned exactly using only signal delay
changes while keeping constant weights in simulations implemented in
Scala (top equation in Figure 1a). We then changed the model to
produce network architectures with a single combined hidden-output
layer, by making the activation function a double-Gaussian and by
summing (or integrating) the squared output from this layer in time
(bottom two equations in Figure 1a). This allows for mutual inhibition
between inputs and solution of the XOR problem. In each model, the
delays are not used to give history to the function and expand the input
space, as is common in time series prediction using ANNs. Instead, by
either explicitly (implementation of top equation of Figure 1a) or
implicitly (bottom two equations in Figure 1a) summing through time, we
allow for the possibility of using delays beyond this input space
expansion use. We show appropriately trained DRNs (bottom equations
of Figure 1a), where all training is on the delay parameters, can
successfully classify on some standard classification datasets. In such a
feedforward context, we see that delay modifications can have a
functional equivalence to weight modifications. We propose such
feedforward delay-using ANNs as biologically more interesting ANNs that
allow for further hypothesis formation and testing of learning and
computational mechanisms in biological neural nets. We show that the
delay space size (equivalently, time to compute a given function) has an
effect on the number of unique bit patterns that can be learned (Figure
1b) and on the complexity of functions that can be modeled (e.g. for
classification). Though the models are still caricatures of biological
networks, they support and suggest a few important ideas: a) that signal
propagation time changes can serve as a mechanism for learning in
biological networks. As per our initial thoughts, these models support the
notion that glial cells might, in vivo, implement such a delay-based
learning mechanism. We believe glia have all the required mechanisms
and couplings for doing this, for example via activity-dependent
myelination modifications [2,3]. Also b) that the maximum waiting-time
for a given region for a signal to arrive from another determines how
sensitive the regions are to inter-region delay changes (e.g. on the axonal
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P214) A shows three different mathematical formulations implemented. Top equation is original one with which exact bit
patterns were learned, as implemented in Scala code. Bottom two equations of A were variations used in more recent work, solving XOR and classifying
data. B shows a plot of amount of bit patterns learned by a 2-3-1 DRN architecture as implemented by the top equation in A, showing the effect of the
maximum allowed delay per link/connection between two nodes (on x-axis) and the number of bits learned (average and maximum) on the y-axis
tracts), and, equivalently, how complex the functional coupling between
the two regions can be and d) that the models show clear tradeoffs
between complexity (or size) of a network, and the time required in
computing a given function. This space-time tradeoff is akin to the many
other tradeoffs and competing optimizations co-existing in the brain [4].
Acknowledgements: We thank Dr. Robert Leech for useful discussions
and comments.
References
1. Hornic K: Multilayer Feedforward Networks are Universal Approximators.
Neural Networks 1989, 2(5):359-366.
2. Fields RD: Myelination: An Overlooked Mechanism of Synaptic Plasticity?
Neuroscientist 2005, 11(6):528-531.
3. Fields RD, Burnstock G: Purinergic signalling in neuron-glia interactions.
Nature Reviews Neuroscience 2006, 7(6):423-436.
4. Bullmore E, Sporns O: The economy of brain network organization. Nature
Reviews Neuroscience 2012, 13: 336-349.
P215
Extracellular potassium concentration defines neuronal bursting properties
Yaroslav I Molkov1*, Bartholomew J Bacak2, Joshua Segaran3, Ilya A Rybak2
1 1. Smith JC, Ellenberger HH, Ballanyi K, Richter DW, Feldman JL: Pre-Botzinger
Department of Mathematical Sciences, Indiana University-Purdue University, IN,
USA; 2Department of Neurobiology and Anatomy, Drexel University College of
Medicine, PA 19123, USA; 3Carmel High School, Carmel, IN 46032, USA
E-mail: ymolkov@iupui.edu
BMC Neuroscience 2015, 16(Suppl 1):P215
Many neurons, or populations of neurons, in the brain are capable of
producing rhythmic bursting activity. This ability is putatively responsible
for rhythmogenic functions like breathing and locomotion. In vivo, rhythms
are generated by synaptically interconnected neuronal networks, whereas
rhythmic bursting behavior is often induced in vitro by elevating the
extracellular potassium concentration (Kout) [1]. It is known that increasing
K out raises the reversal potentials of potassium and leak currents [2].
However, the complete nature of how these effects underlie bursting
activity has yet to be uncovered.
A mathematical modeling study was performed to elucidate the interplay
between these factors and their roles in a neurons transition from
quiescence to rhythmic bursting. A conductance-based model of a
neuron from the pre-Btzinger Complex (pre-BtC) was used as a basis
[3]. A potassium ion component was incorporated into the leak current,
and model behaviors were investigated at varying concentrations of Kout,
taking into account its effect on delayed rectifier potassium current
Page 144 of 200
responsible for after-spike hyperpolarization. The primary aim of this
modeling study was to evaluate the contribution of extracellular
potassium ions in the leak and delayed rectifier potassium current, and
the subsequent effect of these altered currents on the bursting properties
of neurons. Furthermore, the initial model was modified to replicate
experimental results and test for conditions of low Kout as seen in vivo.
The analysis of our model shows that: (i) in vitro bursting behavior with
elevated K out may occur due to attenuation of the delayed rectifier
potassium current and (ii) no oscillations are generated at physiological
levels of extracellular potassium. These results indicate that, according to
the commonly-accepted models used in our study, neurons that naturally
burst in in vitro preparations may not be able to burst in vivo under any
circumstances. Accordingly, rhythmic activity in vivo should rely on other
mechanisms. For example, Jasinski et al. [4] have shown that the
recurrent synaptic excitation in combination with the sodium-potassium
exchanger (pump) can result in the robust rhythmic network activity even
with all intrinsic bursting mechanisms blocked.
Acknowledgements: This work was supported by National Institutes of
Health, grant R01 AT008632 to Y. I. M.; grants R33 HL087377; R01 NS057815;
and R01 NS069220 to I. A. R.
References
Complex - a Brain-Stem Region That May Generate Respiratory Rhythm
in Mammals. Science 1991, 254(5032):726-729.
2. Rybak IA, Shevtsova NA, St-John WM, Paton JFR, Pierrefiche O: Endogenous
rhythm generation in the pre-Botzinger complex and ionic currents:
modelling and in vitro studies. Eur J Neurosci 2003, 18(2):239-257.
3. Butera RJ, Rinzel J, Smith JC: Models of respiratory rhythm generation in
the pre-Botzinger complex. I. Bursting pacemaker neurons. J Neurophysiol
1999, 82(1):382-397.
4. Jasinski PE, Molkov YI, Shevtsova NA, Smith JC, Rybak IA: Sodium and
calcium mechanisms of rhythmic bursting in excitatory neural networks
of the pre-Botzinger complex: a computational modelling study. Eur J
Neurosci 2013, 37(2):212-230.
P216
The suppression curve as a new representation of the premature EEG
maturation
Ninah Koolen1,2*, Anneleen Dereymaeker3, Katrien Jansen3, Jan Vervisch3,
Vladimir Matic1,2, Maarten De Vos4,5, Gunnar Naulaers3, Sabine Van Huffel1,2
1
Division STADIUS, Department of Electrical Engineering (ESAT), University of
Leuven, Leuven, Belgium; 2iMinds-KU Leuven Medical IT Department,
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Leuven, Belgium; 3Department of Development and Regeneration, University
of Leuven, Leuven, Belgium; 4Department of Psychology, University of
Oldenburg, Oldenburg, Germany; 5Institute of Biomedical Engineering,
Department of Engineering Science, University of Oxford, Oxford, UK
E-mail: ninah.koolen@esat.kuleuven.be
BMC Neuroscience 2015, 16(Suppl 1):P216
Automated analysis of premature electroencephalogram (EEG) for
diagnosis is a crucial step to reduce the workload of neurologists. The
grade of discontinuity gives important information about the maturation
[1]. For normal maturation, the discontinuous pattern gradually evolves
into a more continuous pattern. This means, interburst intervals (IBI),
periods of low activity, become shorter. We have defined the suppression
curve (SC), which is a measure of discontinuity [2] (Figure 1A). All data
for this study were recorded at the Neonatal Intensive Care Unit,
University Hospital Gasthuisberg, Leuven, Belgium. The dataset consisted
of 170 EEG recordings (8 channels, 250 Hz) of 93 preterm infants with a
postmenstrual age (PMA) of 24-40 weeks. Some maturational features are
extracted from the discontinuous periods, like the IBI length and the
synchrony index. However, the SC on itself gives also relevant
information about the maturation. Taking the mean of every SC, we can
find a correlation with the age till 34 weeks PMA (Figure 1B). Few outliers
(abnormal EEG) are excluded. After that age, the patient is called late
preterm or even term, and the EEG pattern is in normal condition mostly
continuous (low values of the SC).
In conclusion, this research adds another valuable feature for the
automated analysis of premature background EEG, which would improve
the overall assessment in the NICU for EEG diagnosis.
Acknowledgements: Research Council KUL: GOA/10/09 MaNet, CoE PFV/
10/002 (OPTEC); PhD/Postdoc grants; Flemish Government: FWO, IWT:
projects: TBM 110697-NeoGuard; PhD/Postdoc grants; Belgian Federal
Science Policy Office: IUAP P7/19/ (DYSCO); EU: ERC Advanced Grant:
BIOTENSORS (n 339804).
Figure 1(abstract P216) A Suppression curve example, containing 2 periods of 20-30 minutes of discontinuous pattern, B Evolution of the mean of the
suppression curve in function of the age, represents a patient with normal EEG, * patient with abnormal EEG
Page 145 of 200
References
1. Andr M, Lamblin MD, dAllest AM, Curzi-Dascalova L, Moussalli-
Salefranque F, Nguyen S, et al: Electroencephalography in premature and
full-term infants. Developmental features and glossary. Clin Neurophysiol
2010, 40(2):59-124.
2. Koolen N, Jansen K, Vervisch J, Matic V, De Vos M, Naulaers G, Van Huffel S:
Line length as a robust method to detect high-activity events:
Automated burst detection in premature EEG recordings. Clin
Neurophysiol 2014, 125(10):1985-1994.
P217
Multi-scale detection of rate and variance changes in neuronal spike
trains
Stefan Albert1*, Michael Messer1, Brian Rummell2, Torfi Sigurdsson2,
Gaby Schneider1
1
Institute of Mathematics, Goethe-University, Frankfurt am Main, Germany;
2
Institute of Neurophysiology, Goethe-University, Frankfurt am Main,
Germany
E-mail: albert@math.uni-frankfurt.de
BMC Neuroscience 2015, 16(Suppl 1):P217
Neuronal spike trains can show variability with respect to process
parameters such as the rate or variability of inter spike intervals. These
changes can occur on fast and slow time scales, including also simultaneous
and separate changes in different process parameters.
Building up on results of [1] we present a multiple filter technique (MFT)
that detects change points in the rate and variance of point processes on
multiple time scales simultaneously. In particular, we use a filtered
derivative process and its limit behavior under stationarity. The method
also extends to higher order moments.
The separate detection of rate and variance changes requires two
techniques: First, rate changes need to be detected, irrespective of potential
variance changes. To this end, our approach allows the identification of rate
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
changes in point processes with a certain variability in their lifetimes.
Second, the identified rate changes need to be considered when analyzing
variance changes. We investigate the empirical properties of our asymptotic
MFT method in simulations and apply the MFT to spike trains recorded from
auditory cortex of behaving mice, illustrating rate and variability dynamics
during the task.
Acknowledgements: This work was supported by the German Federal
Ministry of Education and Research (BMBF, Funding number: 01ZX1404B;
SA, TS, GS) and by the Priority Program 1665 of the DFG (TS, GS).
Reference
1. Messer M, Kirchner M, Schiemann J, Roeper J, Neininger R, Schneider G: A
multiple filter test for the detection of rate changes in renewal
processes with varying variance. Annals of Applied Statistics 2014,
8(4):2027-2067.
P218
Joint pausiness in parallel spike trains
Matthias Grtner1*, Sevil Duvarci2, Jochen Roeper2, Gaby Schneider1
1
Institute for Mathematics, Goethe-University, Frankfurt, Germany;
2
Neuroscience Center, Institute of Neurophysiology, Goethe-University,
Frankfurt, Germany
E-mail: gaertner@math.uni-frankfurt.de
BMC Neuroscience 2015, 16(Suppl 1):P218
So-called pauses, i.e., periods with surprisingly few spikes, have recently
gained increasing attention in the analysis of parallel spike trains of
dopaminergic (DA) and Purkinje cells, in particular concerning simultaneity
of pausing activity. The analysis of simultaneous pauses is usually based on
the pauses identified in the separate spike trains. As a consequence, such
techniques can suffer from the local definition of a pause within one spike
train and can thus fail to identify joint pauses across spike trains that are
easily detectable by eye. In addition, they crucially depend on the
algorithm used for pause detection.
In order to tackle this problem, we present a new statistical method for the
detection of synchronous pauses that focuses on typical characteristics of
time periods showing synchronous pauses in parallel spike trains, and
introduce a new measure for synchronous pausiness in parallel spike
trains. We apply the technique to a data set of parallel DA neurons
recorded from the VTA in freely moving mice. Interestingly, pausiness can
be significantly increased in parallel spike trains as compared to individual
processes or processes shifted by small time lags. This observation is
robust and practically independent from the algorithm used for pause
detection.
Acknowledgements: This work was supported by the Priority Program
1665 of the DFG (SD, MG, JR, GS) and by the German Federal Ministry of
Education and Research (BMBF, Funding number: 01ZX1404B; GS).
P219
Temperature-induced changes of spike timing precision and network
synchronisation
Jan-Hendrik Schleimer1,2*, Janina Hesse1,2, Susanne Schreiber1,2
1
Institute for Theoretical Biology, Institute for Biology, Humboldt University,
Berlin, Germany; 2Bernstein Centre for Computational Neuroscience, Berlin,
Germany
E-mail: jh.schleimer@hu-berlin.de
BMC Neuroscience 2015, 16(Suppl 1):P219
Changes in the temperature of neuronal tissue influence electrical activity
in non-trivial ways [1]. Despite a central regulation of body temperature in
mammals, small changes in temperature can have substantial effects. For
example, increases in brain temperature can in rare cases even induce
epileptic seizures [2]. At the level of single cells, temperature affects ion
channels: an increase in temperature predominantly speeds up their
opening and closing rates and increases their maximal conductances.
While the temperature dependence of nervous system function is widely
acknowledged, we currently still lack a full understanding of the effects of
temperature on local circuits.
Here, we investigate the impact of temperature on spike timing precision
and network synchronisation. It is known that properties of individual ion
channel types can modify spiking regularity of single neurons and
Page 146 of 200
network synchronisation. To study temperature-induced changes of these
observables, multiple ion channels have to be considered simultaneously.
Our approach is based on analytical simplification and numerical
continuation. Spike jitter is analysed in temperature-dependent
conductance-based models with ion channel stochasticity. Below
threshold, the dynamics are appropriately described by a coloured noise
escape problem, and, above threshold, by a phase-oscillator description.
We show that for common, physiological parameter combinations, an
increase in temperature reduces spike jitter. Below threshold, this can be
explained by a temperature-induced redistribution of current noise power
to higher frequencies, which are filtered out by the membrane impedance.
Above threshold, the faster dynamics at elevated temperatures alter the
ion channels phase response curves and hence their phase susceptibility.
These results are interpreted in the context of network synchronisation by
showing the effect of temperature on the entrainment region of
deterministic models.
Acknowledgement: This work was funded by grants from the Federal
Ministry of Education and Research, Germany (01GQ1001A, 01GQ0901,
01GQ0972, 01GQ1403) and the Deutsche Forschungsgemeinschaft
(SFB618, GRK1589/1).
References
1. Reig R, Mattia M, Compte A, Belmonte C, Sanchez-Vives MV : Temperature
Modulation of Slow and Fast Cortical Rhythms. Journal of Neurophysiology
2010, 103(3):1253-1261.
2. Sun L, Gilligan J, Staber C, Schutte RJ, Nguyen V, ODowd DK, Reenan R: A
knock-in model of human epilepsy in Drosophila reveals a novel cellular
mechanism associated with heat-induced seizure. J Neurosci 2012,
32(41):14145-14155.
P220
Dendritic distribution of synaptic input creates a trade-off between
input selectivity and flexibility
Michiel Remme*, Susanne Schreiber
Institute for Theoretical Biology, Humboldt University, Berlin, 10115, Germany
E-mail: michiel.remme@hu-berlin.de
BMC Neuroscience 2015, 16(Suppl 1):P220
Information processing by cortical pyramidal neurons is shaped by the
spatial distribution of synapses across the dendrites. A prominent
hypothesis is that synapses with similar selectivities cluster on dendritic
branches. This enables cooperative interactions between neighboring
synapses through activation of voltage-dependent membrane currents,
and helps establish independent integrative subunits, thereby expanding
the computational power of a single neuron [1]. Some recent in vivo
recordings argue against this hypothesis, suggesting that inputs that share
the same stimulus selectivity are randomly distributed throughout the
dendritic tree (e.g., [2]). Other studies seem to support clustered
configurations of input selectivities, showing for example that the activity
of a synaptic input is more strongly correlated with its neighbors than with
more distant inputs [3].
One fundamental feature of the nervous system that has received little
attention in this ongoing discussion is learning; specifically, how is the
ability to change a neurons selectivity (i.e., its flexibility) affected by the
spatial distribution of synapses? This is highly relevant because the
selectivity of many cortical pyramidal neurons is subject to ongoing
modification, not only during development, but throughout adulthood
(e.g., [4]).
Here, we show that the distribution of synapses across active dendrites
shapes both the stimulus selectivity of the neuron, as well as the flexibility
of the neurons selectivity. Using cable theoretic analysis and numerical
simulations of detailed neuron models we show that synapses that are
randomly distributed across the dendrites allow for a modest stimulus
selectivity that can be flexibly modified through synaptic plasticity. In
contrast, a clustered distribution of synapses that encode the same
stimulus allows for very strong stimulus selectivity, however, it hinders
adjustment of this selectivity through synaptic plasticity, which requires
slow and metabolically costly rearrangement of synaptic projections.
Hence, the distribution of synapses across active dendrites creates a trade-
off between selectivity and plasticity. We suggest that pyramidal neurons
with different functions in the cortical information processing hierarchy
exploit different ends of this spectrum.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Acknowledgements: This work was supported by BMBF grants no.
01GQ1001A, 01GQ0901, 01GQ0972, and the Einstein Stiftung Berlin.
References
1. Larkum ME, Nevian T: Synaptic clustering by dendritic signalling
mechanisms. Curr Opin Neurobiol 2008, 18(3):321-331.
2. Jia H, Rochefort NL, Chen X, Konnerth A: Dendritic organization of sensory
input to cortical neurons in vivo. Nature 2010, 464:1307-1312.
3. Takahashi N et al: Locally synchronized synaptic inputs. Science 2012,
335(6066):353-356.
4. Frgnac Y, Shulz D, Thorpe S, Bienenstock E: A cellular analogue of visual
cortical plasticity. Nature 1988, 333:367-370.
P221
Self-organization of information processing in developing neuronal
networks
V Priesemann1,2*, J Lizier3, M Wibral4, ET Bullmore5,6,7, O Paulsen8,
P Charlesworth8, MS Schrter5
1
Department of Nonlinear Dynamics, Max Planck Institute for Dynamics and
Self-Organization, Gttingen, Germany; 2Bernstein Center for Computational
Neuroscience, Gttingen, Germany; 3School of Civil Engineering, University
of Sydney, Sydney, Australia; 4MEG Unit, Brain Imaging Center, Goethe
University, Frankfurt am Main, Germany; 5Behavioural & Clinical Neuroscience
Institute, Department of Psychiatry, University of Cambridge, Cambridge CB2
3EB, UK; 6Cambridgeshire and Peterborough NHS Foundation Trust,
Cambridge CB21 5HH, UK; 7GlaxoSmithKline, Immuno Psychiatry, Alternative
Discovery and Development, Stevenage SG1 2NY, UK; 8Department of
Physiology, Development and Neuroscience, University of Cambridge,
Physiological Laboratory, Downing Street, Cambridge CB2 3EG, UK
E-mail: viola@nld.ds.mpg.de
BMC Neuroscience 2015, 16(Suppl 1):P221
Human brains possess sophisticated information processing capabilities,
which rely on the coordinated interplay of billions of neurons. Despite
recent advances in characterizing the collective neuronal dynamics,
however, it remains a major challenge to understand the principles of how
functional neuronal networks develop and maintain these processing
capabilities. A popular hypothesis is that neuronal networks self-organize
to a critical state [1-3], because in models, criticality maximizes information
processing capacities [4-6]. This predicts that biological networks should
develop towards a critical state during maturation, and at the same time
processing capabilities should increase. We tested this hypothesis using
multi-electrode spike recordings in mouse hippocampal and cortical
neurons over the first four weeks in vitro. We showed that developing
neuronal networks indeed increased their information processing
capacities, as quantified by transfer entropy and active information storage
[6-8]. The increase in processing capacity was tightly linked to decreasing
the distance to criticality (correlation r = 0.68, p < 10-9; r = 0.55, p < 10-6 for
transfer and storage, respectively). Thereby our results for the first time
demonstrate experimentally that approaching criticality with maturation
goes in hand with increasing processing capabilities.
References
1. Beggs JM, Plenz D: Neuronal avalanches in neocortical circuits. J Neurosci
2003, 23(35):11167-11177.
2. Priesemann V, Valderrama M, Wibral M, Le Van Quyen M: Neuronal
avalanches differ from wakefulness to deep sleep-evidence from
intracranial depth recordings in humans. PLoS Comput Biol 2013, 9(3):
e1002985.
3. Priesemann V, Wibral M, Valderrama M, Proepper R, Le Van Quyen M,
Geisel T, Triesch J, Nikolic D, Munk MHJ: Spike avalanches in vivo suggest
a driven, slightly subcritical brain state. Front Syst Neurosci 2014, 8:108.
4. Bertschinger N, Natschlger T: Real-time computation at the edge of
chaos in recurrent neural networks. Neural Comput 2004, 16(7):1413-1436.
5. Boedecker J, Obst O, Lizier JT, Mayer NM, Asada M: Information processing
in echo state networks at the edge of chaos. Theory Biosci 2012,
131(3):205-213.
6. Wibral M, Lizier J, Priesemann V: Bits from Brains for Biologically-Inspired
Computing. Computational Intelligence 2015, 2:5.
7. Lizier JT, Prokopenko M, Zomaya AY: The Information Dynamics of Phase
Transitions in Random Boolean Networks. ALIFE 2008, 374-381.
8. Schreiber T: Measuring Information Transfer. Phys Rev Lett 2000,
85:461-464.
Page 147 of 200
P222
Markov Stability partitioning shows spectrally dependent community
structure amongst thalamocortical neural ensembles
Christian-David Martin*, Silvia Ardila-Jimenez, Simon Schultz
Centre for Neurotechnology & Department of Bioengineering, Imperial
College London, London, UK
E-mail: c.martin14@imperial.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P222
The processing of information through the spatiotemporal coordination of
neuronal activity is still poorly understood [1]. Here we analyse local field
potential (LFP) signals from multi-electrode recordings in the mouse lateral
geniculate nucleus (LGN) and visual cortex (V1), to systematically
investigate interactions between neuronal ensembles across the frequency
spectrum. Computing mutual information for each pair of electrodes using
the k-nearest neighbor method developed by [2], two broad groupings
can be discerned among the electrodes (Figure 1A). The same partitioning
is found as a stable solution when applying the Markov Stability algorithm
developed by Billeh et. al. [3], which uses a Markov diffusion process
through the dataset to detect stable groupings (Figure 1B/C). Analysing
narrowband filtered LFP signals, neuronal groupings were found to change
between low (1-40 Hz) and high (>40Hz) frequency bands. One particular
neural ensemble was found to participate in different groupings across low
and high frequency bands, with differing interaction partners and
mechanisms as assessed by phase-phase and phase-amplitude correlation
measures, both within and across areas. This frequency-specific interaction
pattern may allow for the simultaneous coordination of information
transmission across different timescales.
References
1. Aru J, Aru J, Priesemann V, Wibral M, Lana L, Pipa G, Singer W, Vicente R:
Untangling cross-frequency coupling in neuroscience. Current Opinion in
Neurobiology 2015, 31:51-61.
2. Kraskov A, Stoegbauer H, Grassberger P: Estimating Mutual Information.
Phys Rev E 2004, 69(6):1-16.
3. Billeh YN, Schaub MT, Anastassiou CA, Barahona M, Koch C: Revealing cell
assemblies at multiple levels of granularity. Journal of Neuroscience
Methods 2014, 236:92-106.
P223
A robust model of sensory tuning using dendritic non-linearities
Romain D Caz*, Sarah Jarvis, Simon R Schultz
Centre for Neurotechnology & Department of Bioengineering, Imperial
College London, London, UK
E-mail: r.caze@imperial.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P223
Dendrites, like neurons, can preferentially activate for certain stimuli, but
recent experimental evidence suggests that dendritic tuning can differ
from the neuronal tuning. For instance, in a L2/3 pyramidal neuron in the
mouse visual cortex, dendritic calcium signals display a wide range of
tuning profiles, some of which differ from the tuning of the neuronal
output [1]. This puzzling observation was unanticipated by the standard
Hubel and Wiesel model explaining the origin of visual tuning [2]. The
standard model can survive this observation, but only with the addition of
superfluous synapses. We propose here an alternative model where
synapses responsible for neuronal tuning are dispersed over dendrites.
This alternative model builds on previously published results [3]. It
possesses non-linear dendritic compartments, and in each compartment
the result of multiple excitatory inputs can be smaller than their arithmetic
sum. These non-linear and independent sites of synaptic integration create
neuronal tuning: groups of correlated presynaptic inputs encode the
stimulus identity, and only the group that encodes the preferred stimulus
targets different dendrites and leads to a response. Groups coding for non-
preferred stimuli instead target the same dendrite, and explain the wide
range of dendritic tunings observed experimentally. Moreover, we
demonstrate that this implementation of neuronal tuning is robust to the
loss of dendrites. Thus, our alternative model not only reproduces the
experimental observations, but is also robust to dendritic reorganization.
To confirm this result in silico we use a multi-compartmental model with
a realistic L2/3 morphology [1] (see Figure 1). Our work implies that
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 148 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

Figure 1(abstract P222) A. Mutual Information (MI) between the LFP time-series of each of 32 electrodes within mouse LGN. MI is computed with
entropy estimates from k-nearest neighbor distances [2]. The diagonal was excluded from analysis. Two broad groupings can be discerned in this
example. B. Application of the Markov Stability algorithm [3] to the time-series data in A. The number of communities detected is depicted as a function
of Markov time (blue), with plateau phases indicative of stable partitioning. Stability (orange) of a given partition is defined as the probability that a
random walker at stationarity starts in community i and ends up in the same community after time tM , minus the probability of this happening by
chance, summed over all communities and nodes. C. Variation of information (VI), an information-theoretic measure of the distance between partitions,
averaged over multiple runs of the algorithm (>100x). Low values are indicative of reliable partitioning. Partitioning into two and three groups is relatively
stable across Markov time, and coincides with low VI, thus reflecting a meaningful partitioning

Figure 1(abstract P223) Representation of the biophysical model used in the simulation. Red dots indicate the 7 input sites. The black dot indicates
the soma. 1B. Polar plot showing the tuning of the membrane voltage within two dendritic compartments of our model. 1C. Somatic voltage for three
different stimuli (0, 45, 90 degrees). Red indicates the preferred stimulus, and blue traces non-preferred stimuli

non-linear integration in dendrites can play a pivotal role in neuronal tuning
even if dendritic and neuronal tuning differ. This implementation of
neuronal tuning is also robust to dendritic loss, a property important to
keep a stable sensory representation. Importantly, our theoretical framework
predicts that a neuron is tuned to the input generating the most
widespread synaptic activity on its dendrites and that neuronal tuning can
resist the loss of a significant number of dendrites.
Acknowledgements: This work was supported by EU FP7 Marie Curie
Initial Training Network 289146 NETT . SJ is supported by EU FP7 Marie
Curie fellowship (PIEF-GA-2013-628086).
References
1. Jia H, Rochefort NL, Chen X, Konnerth A: Dendritic organization of sensory
input to cortical neurons in vivo. Nature 2010, 464:1307-1312.
2. Hubel D, Wiesel T: Receptive fields, binocular interaction and functional
architecture in the cats visual cortex. J Physiol 1961, 160(1):106-154.
3. Caz RD, Humphries M, Gutkin B: Passive Dendrites Enable Single
Neurons to Compute Linearly Non-separable Functions. PLoS Comput.
Biol 2013, 9:e1002867.
P224
An information theoretic measure of cross-frequency coupling
Silvia C Ardila-Jimenez1,2*, Jiaying Tang1,2, Simon R Schultz1,2
1
Centre for Neurotechnology, Imperial College London, London, UK;
2
Department of Bioengineering, Imperial College London, London, UK
E-mail: sca11@imperial.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P224
The coupling of neuronal oscillations between cortical areas has been
proposed both as a mechanism for top-down and bottom-up signaling in
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P224) Mutual Information performance compared against other methods. A. Scatter plots of CFC measures using a synthetic data
MVL vs. MI, ESC vs. MI and ESC vs. MVL. Different colors indicate different noise variance. B. Mean and variance of the CFC measures for the different
levels of noise. C. MI used to explore CFC between the thalamus and cortex
the brain. These interactions may facilitate the coordination of both local
and distributed networks across different time scales. However, we are
still exploring what the best method is to quantify them. Of particular
interest has been the role of phase-amplitude cross-frequency coupling
(CFC), and a variety of methods to analyze this have been proposed [1].
Thus far none of these stands out as an ideal measure. We propose the
use of Mutual Information to quantify CFC. We test the performance of
this method against two other approaches: the Mean Vector Length, and
the Envelope to Signal Correlation. Finally we apply this method to data
recorded from the mouse early visual system.
The ability of Mutual Information to measure the amount of common
information shared by the two systems, while capturing both linear and
nonlinear relationships, makes it a suitable candidate as a measure of
cross-frequency interactions. The resulting value is measured in an
absolute scale which provides a framework for comparison across studies.
A number of methods have been proposed for measuring CFC each with
its own benefits and caveats.
We show that the Mutual Information can quantify phase amplitude cross-
frequency interactions on an absolute scale. This method performs at least
as well as other CFC measures in identifying the presence of CFC, while
not being restricted to linear interactions. Furthermore, it is robust to
phase shifts between the low-frequency signal and the amplitude of the
high-frequency signal, making it less prone to produce false negative
results. Applying the method to experimental data shows apparent CFC
interactions between the 1-4Hz band in the LGN and the 10-20Hz and 31-
35Hz bands in V1.
Reference
1. Canolty R, Knight R: The functional role of cross-frequency coupling.
Trends in Cog Sci 2010, 14(11):506-515.
Page 149 of 200
P225
Looking at the role of direct and indirect pathways in basal ganglia
networks at different levels
Rahmi Elibol*, Neslihan Serap engr
Electronics and Communication Engineering, Istanbul Technical University,
Istanbul, Turkey
E-mail: rahmielibol@itu.edu.tr
BMC Neuroscience 2015, 16(Suppl 1):P225
The role of basal ganglia in motor action initiation and selection has been
well studied and now it is evident that impairment in this structure causes
not only causes movement disorders as Parkinsons disease, Huntingtons
disease but also behavioral dysfunctions as addiction, obsessive-
compulsive disorder [1-4]. In order to understand the mechanisms giving
rise to motor actions, cognitive processes related to these actions as
decision making and the diseases occurring due to malfunctioning of
these structures, various computational models of direct and indirect
pathways have been proposed [5-9]. Here, in order to set a simple relation
between models of basal ganglia at different levels, a simple mass model
indicating the controversial role of direct and indirect pathways will be
introduced first. While dopamine (DA) in direct pathway enhances the
activity in Thalamus giving rise to inhibition of action, arise of DA in
indirect pathway disinhibits Thalamus activity promoting the action to take
place. This activity can be followed from Figure 1 for different DA levels.
Based on the results of this simple mass model, spiking neural network
(SNN) is built by point neurons and the relation between the local field
potential of this SNN and simple mass model will be discussed. The aim is
to build a simple relation between different levels of computational
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P225) DA level is normal, high and low at top,
middle and bottom, respectively. THL activity is normal, high and low
similar to STRD1 activity and contrary to STRD2 activity
models which would help investigating the mechanisms behind the
cognitive processes without engaging in detailed models initially. Thus, the
simple mass model proposed would be primary model giving a chance to
test the initial interpretation of the concepts formed and lead to setting up
more detailed, realistic models.
References
1. Kropotov J, Etlinger S: Selection of actions in the basal ganglia-
thalamacortical circuits: Review and model. International Journal of
Psychophysiology 1999, 31(3):197-217.
2. DeLong MR, Wichmann T: Circuits and circuit disorders of the basal
ganglia. Archives of Neurology 2007, 64(1):20-24.
3. Nicola S: The nucleus accumbens as part of a basal ganglia action
selection circuit. Psychopharmacology 2007, 191(3):521-550.
4. Graybiel AM, Rauch SL: Toward a Neurobiology of Obsessive-Compulsive
Disorder. Neuron 2000, 28(2):343-347.
5. Terman D, Rubin JE, Yew AC, Wilson CJ: Activity patterns in a model for
the subthalamopallidal network of the basal ganglia. The Journal of
Neuroscience 2002, 22(7):2963-2976.
6. Chersi F, Mirolli M, Pezzulo G, Baldassarre G: A spiking neuron model of
the cortico-basal ganglia circuits for goal-directed and habitual action
learning. Neural Networks 2013, 41:212-224.
7. McCarthy MM, Moore-Kochlacs C, Gu X, Boyden ES, Han X, Kopell N: Striatal
origin of the pathologic beta oscillations in Parkinsons disease.
Proceedings of the National Academy of Sciences 2011, 108(28):11620-11625.
8. Marreiros AC, Cagnan H, Moran RJ, Friston KJ, Brown P: Basal ganglia
cortical interactions in Parkinsonian patients. NeuroImage 2013,
66:301-310.
Page 150 of 200
9. Yucelgen C, Denizdurduran B, Metin S, Elibol R, Sengor NS: A biophysical
network model displaying the role of basal ganglia pathways in action
selection. Artificial Neural Networks and Machine Learning ICANN 2012,
7552:177-184.
P227
Curvature of dendritic nonlinearities modulates higher-order spiking
correlations
Alex Cayco-Gajic1*, Joel Zylberberg2, Eric Shea-Brown2
1
Department of Neuroscience, Physiology & Pharmacology, University
College London, London, UK; 2Department of Applied Mathematics,
University of Washington, Seattle, WA, USA
E-mail: natasha.gajic@ucl.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P227
Characterizing neural spiking covariability is essential for understanding
the collective activity of neural populations. Recent experiments have
provided evidence of statistical dependencies among groups of neurons
beyond that expected by the firing rates and pairwise correlations alone
[1-3]. These higher-order correlations (HOCs) can be generated by
common input [4] or motifs within the network architecture [5], yet a
complete mechanistic understanding is lacking. We explore a novel
mechanism through which higher-order correlations can be modulated:
dendritic nonlinearities. We simulated the spiking activity of a simple
exponential integrate-and-fire model neuron in response to two correlated
presynaptic spike trains and background noise. The synaptic conductances
were either summed linearly at the soma or were filtered through a
nonlinear dendritic transfer function (see Figure 1), which was chosen to
have a similar magnitude of effect on nonlinear EPSP summation as
observed by active dendritic properties in pyramidal cells [6]. Using
maximum entropy techniques, triplet correlations in the circuit were
quantified as the probability of synchronous triplet spiking between the
postsynaptic cell and its two presynaptic inputs beyond what could be
captured by the firing rates and pairwise correlations alone, normalized by
the product of the standard deviations. We found that superlinear
dendrites significantly increased the level of triplet. On the other hand,
saturating dendrites decorrelated triplets. These results reveal that HOCs in
spiking activity are modulated by the curvature of the dendritic transfer
function. Finally, this study demonstrates how intrinsic single-cell
properties can tune spiking covariability in neural populations.
References
1. Ohiorhenuan IE, Mechler F, Purpura K, Schmid A, Hu Q, Victor JD: Sparse
coding and high-order correlations in fine-scale cortical networks. Nature
2010, 466:617-621.
2. Ganmor E, Segev R, Schneidman E: Sparse low-order interaction network
underlies a highly correlated and learnable neural population code.
PNAS 2011, 108:9679-9684.
3. Koester U, Sohl-Dickstein J, Gray C, Olshausen B: Modeling higher-order
correlations within cortical microcolumns. PLOS Comp Bio 2014,
10:1003684.
4. Macke J, Opper M, Bethge M: Common input explains higher-order
correlations and entropy in a simple model of neural population
activity. Phys Rev Lett 2011, 106.
5. Hu Y, Trousdale J, Josic K, Shea-Brown E: Local paths to global coherence:
Cutting networks down to size. Phys Rev E 2014, 89:032802.
6. Polsky A, Mel B, Schiller J: Computational subunits in thin dendrites of
pyramidal cells. Nature Neurosci 2004, 7:621-627.
P228
Large-scale quantitative analysis of neurons via morphological
structures by Fast Automatically Structural Tracing Algorithm (FAST)
Nan-Yow Chen1*, Kuan-Peng Chen1, Chi-Tin Shih2*, Guan-Wei He3,
Ting-Yuan Wang4, Yu-Tai Ching3, Ann-Shyn Chiang4
1
National Center for High-Performance Computing, Hsinchu 30076, Taiwan,
Republic of China; 2Department of Physics, Tunghai University, Taichung 40704,
Taiwan, Republic of China; 3Department of Computer Science, National Chiao
Tung University, Hsinchu 30010, Taiwan, Republic of China; 4Department of Life
Science, National Tsing Hua University, Hsinchu 30013, Taiwan, Republic of China
E-mail: nanyow@nchc.narl.org.tw
BMC Neuroscience 2015, 16(Suppl 1):P228
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P227) A. Transfer functions depicting the summation of synaptic inputs for linear dendrites. B. Transfer function for saturating
dendritic nonlinearities (solid line), as well as the identity (dotted line). C. Same as B for superlinear dendrites. D-F. Triplet correlations (see text)
between the postsynaptic cell and its presynaptic inputs, shown for D. linear, E. sublinear, and F. superlinear dendrites. Axes represent the varying
statistics of the presynaptic spike trains, i.e., their firing rates and the correlations between them
Quantitative analysis of neurons is a very important issue in neural
science especially after numerous three-dimensional neural images in
Drosophila brains were taken from confocal laser scanning microscope [1].
However, analyzing these messy data is mostly by human being with
some semi-automatic software packages so far. Not only the task is very
labor intensive but also the result is susceptible to errors and usually
lacks objectivity. Therefore, fast and accurate analyzing tools are crucial
and very desirable. Recently, we developed a computational algorithm,
FAST (Fast Automatically Structural Tracing algorithm), which can trace
neurons and get characteristic quantities of neuron fibers from their
morphology in a very efficient way. These characteristic quantities (called
SIs, Structural Indexes) are, for example, number of branch points,
number of end points, cross section area of fibers, branch angle of fibers,
distribution of fiber length, curvature of fibers, and innervation in
neuropils, etc. After structural indexes of neuron fibers were obtained,
isomap [2] and modularity [3] methods are applied to classify neurons
without depending on human intervention. The isomap method can
defined the similarity between neurons by geodesic paths in a high-
dimensional manifold as well as the modularity method can find the best
community structure of classification by optimization, i.e., to maximize
the intra module connections as many as possible and to minimize the
inter module connections as few as possible. With these tools, large-scale
neural morphological structures, their annotations as well as quantified
characteristics, and neural classifications can be facilely and reliably
retrieved as useful data for computational neuroscience.
References
1. Chiang AS, Lin CY, Chuang CC, Chang HM, Hsieh CH, Yeh CW, et al: Three-
Dimensional Reconstruction of Brain-wide Wiring Networks in Drosophila
at Single-Cell Resolution. Current Biology 2010, 21(1):1-11.
2. Tenenbaum JT, de Silva V, Langford JC: A Global Geometric Framework
for Nonlinear Dimensionality Reduction. Science 2000,
290(5500):2319-2323.
3. Guimer R, Amaral LAN: Functional Cartography of Complex Metabolic
Networks. Nature 2005, 433:895-900.
Page 151 of 200
P229
Symmetries constrain the transition to heterogeneous chaos in
balanced networks
Andrea K Barreiro1*, J Nathan Kutz2, Eli Shlizerman2
1
Department of Mathematics, Southern Methodist University, Dallas, TX, USA;
2
Department of Applied Mathematics, University of Washington, Seattle, WA,
USA
E-mail: abarreiro@smu.edu
BMC Neuroscience 2015, 16(Suppl 1):P229
Biological neural circuits display both spontaneous asynchronous activity,
and complex, yet ordered activity while actively responding to input.
Recently, researchers have demonstrated how this spontaneous behavior
underlies computational capabilities in large, recurrently connected
networks of firing rate [1,2] and spiking [3] units.
Yet, not all spontaneous activity is equal: complex computations may
require the rich phase-space of heterogeneous chaos, in which each neuron
has a different time-dependent firing rate [2,3]. Here, we address the
question of how network connectivity structure may affect the transition
to heterogeneous chaos in echo-state networks.
We choose a family of firing-rate networks in which the neurobiological
constraint of Dales Law that most neurons are either excitatory or
inhibitory is satisfied, and in which excitation and inhibition are balanced.
We first study the transition to heterogeneous chaos in this setting, using
principal component analysis (PCA) to provide a lower-dimensional
description of network activity. We find that key characteristics of this
transition differ in constrained networks, versus unconstrained networks
with similar variability: the transition to heterogeneous chaos occurs at
higher coupling strengths, and is variable across specific networks.
These properties are a consequence of the fact that the constrained
system may be described as a perturbation from a system with non-trivial
symmetries. These symmetries imply the presence of both fixed points
and periodic orbits that are an organizing center for solutions, even for
large perturbations. In comparison, spectral characteristics of the network
coupling matrix [4-6] are relatively uninformative about the behavior of
the constrained system.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P228) A schematic diagram for innervation table and classification results of local neurons in olfactory system of Drosophila
We next investigated the impact of this structure on the computational
capabilities of constrained vs. unconstrained networks. We first examine
the response of networks to time-dependent inputs [3]. In comparison to
unconstrained networks, constrained networks performed better on
population coding of firing rate, but had less ability to separate two
different time-dependent inputs in phase space.
In contrast, the delayed transition to chaos had little effect on the ability
of constrained networks to reproduce a learning task recently
Page 152 of 200
investigated in unconstrained networks [7]. We inspected example
networks and found that the addition of the feedback loop quickly
moves effective network connectivity away from symmetry and into the
chaotic regime at the onset of training.
References
1. Bertschinger N, Natschlger T: Real-Time Computation at the Edge of
Chaos in Recurrent Neural Networks. Neural Computation 2004,
16(7):1413-1436.
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 153 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

2. Sussillo D, Abbott LF: Generating Coherent Patterns of Activity from
Chaotic Neural Networks. Neuron 2009, 63(4):544-557.
3. Ostojic S: Two types of asynchronous activity in networks of excitatory
and inhibitory spiking neurons. Nature Neuroscience 2014, 17:594-600.
4. Rajan K, Abbott LF: Eigenvalue Spectra of Random Matrices for Neural
Networks. Physical Review Letters 2006, 97(18):188104.
5. Wei Y: Eigenvalue spectra of asymmetric random matrices for
multicomponent neural networks. Physical Review E 2012, 85:066116.
6. Aljadeff J, Stern M, Sharpee T: Chaos in heterogenous neural networks: I.
The critical transition point. BMC Neuroscience 2014, 15(Suppl 1):O20.
7. Sussillo D, Barak O: Opening the Black Box: Low-Dimensional Dynamics in
High-Dimensional Recurrent Neural Networks. Neural Computation 2013,
25(3):1-24.
P230
Predicting surgical outcome in intractable epilepsy using a
computational model of seizure initiation
Nishant Sinha1*, Justin Dauwels1, Yujiang Wang2, Sydney S Cash3,
Peter N Taylor2
1
School of Electrical and Electronics Engineering, Nanyang Technological
University, Singapore; 2School of Computing Science, Newcastle University,
Newcastle upon Tyne, UK; 3Massachusetts General Hospital and Harvard
Medical School, Boston, MA, USA
E-mail: jdauwels@ntu.edu.sg
BMC Neuroscience 2015, 16(Suppl 1):P230
A third of patients with epilepsy are refractory to anti-epileptic drug
treatment. For some of these patients with focal epilepsy, better seizure
control can be achieved by surgical treatment in which the seizure focus
is localized and resected while avoiding crucial cortical tissues. However,
approximately 30% of the patients continue to have seizures even after
surgery. In other words, reliable criteria for patients outcome prediction
are absent. Computational models with appropriate parameter setting
and patients specific connectivity allows an exciting opportunity to make
predictions based on the model dynamics.
In this study, non-seizure (inter-ictal) epoch of electrographic recording has
been used to calculate the functional synchrony between different cortical
regions. This synchrony measure was then used as the connectivity
parameter in a computational model of transitions to a seizure like state.
Hypothesizing that the network synchrony plays an important role in
determining the likelihood of surgical success, we retrospectively analyzed
19 patients having intractable epilepsy, who underwent surgical treatment
to achieve seizure freedom. All data were collected confirming to ethical
guidelines and under protocols monitored by the local Institutional Review
Boards according to NIH guidelines.
Building upon the computational model in [1], the regions which were
more likely to transit into a seizure like state were delineated. It was found
that these regions are correlated with those identified by clinicians as the
seizure onset zone. Moreover, it was found that the resection of these
regions in the model reduces the overall likelihood of a seizure. The
likelihood of a surgical success was calculated in silico by iteratively
increasing the area of resection and the surgical outcomes were
successfully predicted for 14 out of 19 patients.
The methods presented here may aid clinicians to delineate the seizure
focus. Moreover, it may facilitate neurosurgeons in predicting the
likelihood of a surgical success and to investigate alternative cortical
tissues to operate on if the seizure focus is in the eloquent cortex.
Acknowledgements: This work is funded in part by MOE Academic
Research Funding Tier 1 grant M4010982.040.
Reference
1. Nishant Sinha, Dauwels Justin, Wang Yujiang, Cash SSydney, Taylor NPeter:
An in silico approach for pre-surgical evaluation of an epileptic cortex.
Engineering in Medicine and Biology Society (EMBC), 2014 36th Annual
International Conference of the IEEE 2014, 4884-4887.

Table 1(abstract P230) Prediction of surgical outcomes

No Age group Age group Surgical Resection Outcome Predicted Outcome
at onset at surgery (Engel Class)
1 21-30 21-30 Right Temporal lobe Seizure Free (II) Good outcome
2 41-50 41-50 Right Temporal lobe Seizure Free (I) Good outcome
3 21-30 21-30 Left Cingulate Seizure Free (I) Bad outcome
4 41-50 41-50 Left Temporal Seizure Free (I) Good outcome
5 11-20 11-20 Right Parietal Seizure Free (I) Good outcome
6 51-60 51-60 Amygdalohippocampectomy Seizure Free (I) Bad outcome
Left Medial Frontal Lobe
7 11-20 11-20 Right anterior-superior frontocortical Seizure Free (I) Good outcome
Right Temporal Lobe, Amygdalohippocampectomy
8 11-20 11-20 Left occipital brain lobe Seizure Free (I) Bad outcome
9 31-40 31-40 Right frontal lobe Seizure Free (I) Good outcome
10 1-10 1-10 Left lateral frontal cortex, Seizure Free (I) Bad outcome
Left anterior frontal cortex
Mesial left frontal cortex
11 21-30 21-30 Left FrontoTemporal Not Seizure Free Bad outcome
12 31-40 31-40 Right Temporo-Occipital Region Not Seizure Free (IV) Bad outcome
13 21-30 21-30 Right Temporal Lobe Not Seizure Free (IV) Bad outcome
14 11-20 11-20 Left Anterior Temporal Lobe Not Seizure Free (V) Good outcome
Amygdalohippocampectomy
15 1-10 1-10 Left Parietal Lobe Not Seizure Free (IV) Bad outcome
16 1-10 1-10 Right Frontal Lobe Not Seizure Free (IV) Bad outcome
17 31-40 31-40 Left Temporal Not Seizure Free (V) Bad outcome
18 21-30 21-30 Left Temporal Lobe Not Seizure Free (V) Bad outcome
19 1-10 1-10 Left Frontal Lesion Not Seizure Free (V) Bad outcome
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
P231
EEG slow-wave mediates the fragmentation and coupling of cortical
networks in propofol-induced general anesthesia
Kaier Wang1, Moira L Steyn-Ross1, Alistair Steyn-Ross1*, Marcus T Wilson1,
Jamie W Sleigh2
1
School of Engineering, The University of Waikato, Hamilton, 3240, New
Zealand; 2Waikato Clinical School, The University of Auckland, Waikato
Hospital, Hamilton, 3240, New Zealand
E-mail: asr@waikato.ac.nz
BMC Neuroscience 2015, 16(Suppl 1):P231
Electroencephalogram (EEG) recorded from propofol-induced general
anesthesia is characterized by large amplitude slow-waves (0.11.5 Hz).
Clinically, these lowest frequency components of the EEG signal become
dominant over other higher frequency components during and after loss
of consciousness [1]. However, it remains unclear how these slow
oscillations are produced and to what extent they reflect changes in
cortical network connectivity. Modeling anesthesia as a moderate
reduction in interneuronal gap-junction coupling, a recent theoretical work
by Steyn-Ross et al [2] predicts emergence of anesthetic slow-waves with
chaotic dynamics. In the modeled anesthesia state, the weakened gap-
junction coupling supports a codimension-2 bifurcation point where
competing Turing (space) and Hopf (time) dynamics coexist, signifying
spontaneous symmetry-breaking instabilities in the firing behavior of
cortical neurons. Further, these chaotic slow-waves are found to perturb
the neuronal coupling across the cortex, leading to a dramatic drop in
global phase-coherence compared to its high level during consciousness.
In this study, we analyze clinically-recorded EEG data to examine the
model prediction for changes in phase-coherence between pairs of EEG
channels in the sub-delta band during propofol anesthetic induction. Our
study finds a coherence decrease in the frontal and occipital regions (see
left panel of Figure 1), as well in the connection between them.
Concomitantly, more strongly coupled neuronal activities are disclosed in
the temporal-frontal, temporal-occipital and left-right temporal regions
(right panel). Our clinical observation of reduced EEG coherence is
Figure 1(abstract P231) Correlation representations showing electrode pairs with significantly reduced (left, blue) or increased (right, pink)
phase-coherence for sub-delta band (0.05-1.5 Hz) EEG induced by propofol anesthesia. Only electrode pairs (from 128 scalp electrodes) showing
significant (p < 0.025) change in phase coherence are connected with lines
Page 154 of 200
consonant with cortical model predictions. However, our EEG study
indicates that the coherence alternation is regional in nature, while the
cortical model describes a spatially-uniform trend. Moreover, we did not
find any theoretical prediction for the left- and right-temporal increased-
coherence patterns. As the cortical model by Steyn-Ross et al. is spatially
homogenous, i.e., there are no explicit front-to-back or right-to-left
directionality, it is unable to produce regional coherence changes. It
appears that the Steyn-Ross cortical model best represents the cortical
dynamics in the frontal region.
References
1. Lewis LD, Weiner VS, Mukamel EA, Donoghue JA, Eskandar EN, Madsen JR,
et al: Rapid fragmentation of neuronal networks at the onset of
propofol-induced unconsciousness. Proc. Natl. Acad. Sci. U.S.A 2012, 109:
E3377-E3386.
2. Steyn-Ross ML, Steyn-Ross DA, Sleigh JW: Interacting Turing-Hopf
instabilities drive symmetry-breaking transitions in a mean-field model
of the cortex: a mechanism for the slow oscillation. Phys. Rev. X 2013,
3:021005.
P232
Thalamo-cortical mechanisms of the observed specific changes in
frontal and occipital EEG rhythms during propofol-induce sedation
Meysam Hashemi1*, Axel Hutt1, Jamie Sleigh2
1
I NRIA CR Nancy - Grand Est, Villers-les-Nancy, France; 2Department of
Anaesthetics, Waikato Hospital, Hamilton, New Zealand
E-mail: meysam.hashemi@inria.fr
BMC Neuroscience 2015, 16(Suppl 1):P232
Although general anesthesia is widely used in todays medical surgery, its
precise underling mechanism is not yet clear. For clinically relevant
concentration of propofol specific changes in electroencephalogram (EEG)
rhythms can be observed experimentally. These characteristic changes
comprised increased activity in the delta (0.5-4) Hz and alpha (8-13) Hz
frequency bands over the frontal head region, but increased delta and
decreased alpha activity over the occipital region [1]. The work model aims
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 155 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

3. Hashemi M, Hutt A, Sleigh J: Anesthetic action on extra-synaptic

receptors: effects in neural population models of EEG activity. Front Syst
Neurosci 2014, 8:232.

P233
Description and removal of background activity in EEG power spectra
under general anesthesia using the Lorentzian curve
Mariia Fedotenkova1,2,3*, Axel Hutt1,2,3, James W Sleigh4
1
CNRS, Loria, UMR n 7503, Vanduvre-ls-Nancy, F-54500, France;
2
NEUROSYS team, Inria, Villers-ls-Nancy, F-54600, France; 3Universit de
Lorraine, Loria, UMR n 7503, Vanduvre-ls-Nancy, France; 4Department of
Anesthesia, Waikato Clinical School of the University of Auckland, Waikato
Hospital, Hamilton 3206, New Zealand
E-mail: mariia.fedotenkova@inria.fr
BMC Neuroscience 2015, 16(Suppl 1):P233

General anesthesia is an important medical procedure in todays hospital

practice and comprises loss of consciousness, analgesia, amnesia and
immobility. Our current work analyzes patient reaction on nociception
stimuli during a surgical operation and differences in this reaction
provided different anesthetic drugs, propofol and desflurane. The studied
dataset comprises EEG-recordings before and after incision obtained from
115 patients [1]. The task is the identification of spectral EEG signal
features reflecting the incision. This analysis will reveal a possible new
marker of pain during deep anesthesia.
This work considers one of the approaches to the problem, namely spectral
analysis. First, power spectral density (PSD) estimates were obtained using
Figure 1(abstract P232) Schematic of a thalamo-cortical module. Welchs method. It is well known that EEG power spectrum decays with
The blue arrows indicate excitatory connections and the red arrows higher frequencies following ~1/f scaling [2-4]. We attribute this behavior
represent inhibitory connections. The symbols E, I, S and R denote the to background activity [5], which takes place in the brain when no other
cortical excitatory and inhibitory neurons, thalamo-cortical relay, and activity is present. Background activity was describe by fitting regression
thalamic reticular neurons, respectively. In addition the connections curve P(f)~a/f b to each PSD estimate. However, the resulting goodness of
between cortex and thalamus are associated with the same time delay fit was not satisfactory. It is due to rise of power in delta peak, which
becomes prominent under general anesthesia and makes the process of
curve fitting less reliable. Thus, the original model was substituted by the
Lorentzian function P(f)=a/(f b + c), which resembles the shape of actual
to understand the mechanisms underlying these specific changes in EEG power spectrum quite well. Afterwards, regression curves were subtracted
from each power spectrum to normalize it [3] and to analyze spectral
power spectrum using a neuronal population model of a single thalamo-
power contained in delta and alpha peaks regardless of distinctions in
cortical module (Figure 1) based on a recently developed neural field
patients.
model of anesthetic action [2,3]. The model reproduces well the certain
changes observed experimentally in EEG rhythms over both frontal and
occipital electrodes during propofol anesthesia sedation.
The power spectral analyses reveal that the alpha power originates from
the cortico-thalamic relay interaction, which is associated with a constant
time delay around the inverse of peak frequency in alpha band. It is
shown that as the concentration of propofol increases, dependent on the
potential values of the resting state of the system, it causes an increase
or decrease in the gains function within the thalamo-cortical loop what
then results in an increase or decrease in the spectral power in the alpha
band over frontal and occipital regions, respectively. The model indicates
the importance of multiple resting states in brain activity. Moreover our
findings demonstrate that the emergence of delta power results from the
increased GABAergic inhibition into the thalamo-cortical system.
Our results reveal that the specific observed changes in EEG rhythms can
be reproduced with and without the propofol effect in cortical cells. This
finding points out the importance of thalamus for neural effects under
anesthesia sedation and simplifies the model under study. By reducing
the dimensionality of the model we are able to obtain some inequality
conditions for the stability of the system. In addition, the analytical
tractability of the model allows us to obtain further insight into the
mechanisms underlying the characteristic spectral features seen during
anesthesia sedation.
References
1. Cimenser A, Purdon PL, Pierce ET, Walsh JL, Salazar-Gomez AF, Harrell PG,
et al: Tracking brain states under general anesthesia by using global
coherence analysis. PNAS 2011, 108(21):8832-8837. Figure 1(abstract P233) Example of power spectrum density estimate
2. Hutt A: The anaesthetic propofol shifts the frequency of maximum (solid black), with fitted Lorentzian curve (dashed orange), goodness
spectral power in EEG during general anaesthesia: analytical insights of fit (R2), delta (green circle) and alpha peaks (blue circle)
from a linear model. Front Comput Neurosci 2013, 7:2.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
The results of this work revealed small differences between propofol and
desflurane. Power spectra of patients receiving desflurane have more
regular shape than the ones from propofol group. It can also be seen that
delta power remains more consistent, while alpha power varies greatly
from patient to patient. Another result of this work is a trend in the
distribution of Lorentzian curve parameters: the set of parameters remains
compact for small values of b, but a and c scatters significantly when b
(which corresponds to steepness of curve) is larger than three. Results of
this work provide insights on underlying background activity. However,
they do not allow to distinguish between pre- and post-incision and poorly
between propofol and desflurane. This problem requires more complicated
techniques. Future work will expand spectral analysis with time
information (time-frequency representations) and investigate time
structure by means of recurrence analysis.
References
1. Sleigh JW, Leslie K, Voss L: The effect of skin incision on the
electroencephalogram during general anesthesia maintained with
propofol or desflurane. Journal of Clinical Monitoring and Computing 2010,
24(4):307-318.
2. Bdard C, Destexhe A: Macroscopic Models of Local Field Potentials and
the Apparent 1/f Noise in Brain Activity. Biophysical Journal 2009,
96(7):2589-2603.
3. Hutt A, Frank TD: Critical fluctuations and 1/f a-activity of neural fields
involving transmission delays. Acta Phys Pol A 2005, 108:1021-1040.
4. Demanuele C, James CJ, Sonuga-Barke EJ: Distinguishing low frequency
oscillations within the 1/f spectral behaviour of electromagnetic brain
signals. Behavioral and Brain Functions 2007, 3:62.
5. Freeman WJ, Zhai J: Simulated power spectral density (PSD) of
background electrocorticogram (ECoG). Cognitive Neurodynamics 2009,
3(1):97-103.
P234
Functional requirements for homeostatic inhibitory plasticity in
recurrent networks
Owen Mackwood1,2*, Henning Sprekeler1,2
1
Technische Universitt Berlin, 10587 Berlin, Germany; 2Bernstein Center for
Computational Neuroscience, 10115 Berlin, Germany
E-mail: owen.mackwood@bccn-berlin.de
BMC Neuroscience 2015, 16(Suppl 1):P234
Homeostatic regulation of neuronal firing rates has been reliably observed
in response to chronic manipulation of neural activity. From such
experiments a variety of putative homeostatic mechanisms have been
reported, including compensatory changes for several synapse types [1].
Unfortunately, it is sometimes unclear whether these synaptic changes
occur in response to persistent rate deviations in individual neurons (local
control) or in larger populations (global control). One variant of the former
type is the recently proposed Hebbian learning rule of Vogels et al., which
modifies inhibitory-to-excitatory synapses to maintain a target rate in each
postsynaptic cell [2]. Since this rule asserts its control by changing
inhibition, one might expect that in recurrent networks, similar synaptic
plasticity rules elsewhere within the inhibitory feedback loop should be
capable of comparable homeostatic control. Using simulations of locally
connected recurrent networks, we investigate a plasticity rule that changes
excitatory-to-inhibitory synapses such that presynaptic excitatory activity
should remain at a target rate. We show that if the rule attempts to control
excitatory neurons individually, it induces competition between those cells.
For a Hebbian rule, this effect results in most of the network being driven
to quiescence, whereas with rule dependent only upon presynaptic rates
the competition in less pronounced. In either case, this competition is a
consequence of having a local sensor (measuring the rate of a single cell),
but actuators (synapses) with non-local effect synaptic changes triggered
by an individual excitatory cell also affect other excitatory cells due to the
fanout in recurrent connectivity. Said competition can be detected in the
spontaneous firing rates of the population, which become sparser over
time. If instead the synaptic learning rule attempts to control the average
firing rate of neurons in a sufficiently large region, this competitive effect is
mitigated. Our results suggest that an apparently homeostatic rule can
lead to competition, whose degree depends upon the spatial specificity of
its firing rate sensor and the spatial range of its effect on the network (the
effector). This interaction between sensor and effector locality in
Page 156 of 200
homeostatic rate control implies that further insights into the synaptic
mechanisms of homeostasis could be gained by studying neuronal
responses to localized experimental manipulations.
Acknowledgements: This work was supported by the BMBF through a
Bernstein Award to Henning Sprekeler (FKZ 01GQ1201)
References
1. Turrigiano G: Too Many Cooks? Intrinsic and Synaptic Homeostatic
Mechanisms in Cortical Circuit Refinement. Annu Rev Neurosci 2011,
34:89-103.
2. Vogels TP, Sprekeler H, Zenke F, Clopath C, Gerstner W: Inhibitory plasticity
balances excitation and inhibition in sensory pathways and memory
networks. Science 2011, 334(6062):1569-1573.
P235
Structural plasticity and associative memory in balanced neural
networks with spike-time dependent inhibitory plasticity
Ankur Sinha*, Neil Davey, Roderick Adams, Volker Steuber
Science and Technology Research Institute, University of Hertfordshire,
Hatfield, UK
E-mail: a.sinha2@herts.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P235
In recent work, Vogels and collaborators demonstrated the ability of
spike-time dependent inhibitory plasticity to stabilise recurrent spiking
neural networks by balancing out the excitatory input received by
neurons in the network with the required amount of inhibition [1].
Further, as an application of this unsupervised balance, they showed that
such a network can operate as a non-attractor associative memory, by
being able to store patterns that become indistinguishable from the
background activity of the network, but that can be recalled successfully
by external stimuli.
In other research, also related to homoeostasis in neuronal networks, Butz
and colleagues investigated the rewiring of deafferentated neurons driven
by homoeostatic structural plasticity [2,3]. Their work, based on biological
observations of restructuring in the visual cortex [4,5], suggests that
neurons endeavour to maintain an appropriate level of electrical activity
by adjusting the number of synaptic contacts they make. In order to
remedy the loss in electrical activity experienced by neurons as a result of
deafferentation, they implemented a growth rule that enabled the neurons
to increase the number of lateral connections.
In our research, we measure the memory capacity of balanced spiking
neural networks possessing capabilities of both fast synaptic and slow
structural plasticity by calculating a signal to noise ratio [6]. The memory
capacity of a network of neurons is dependent on various parameters such
as the magnitude of the recall stimulus, the size of the pattern relative to
the size of the network, the learning rule that dictates inhibitory plasticity,
and so on. In this work, as an initial step, we investigate the memory
capacity of a non lesioned network containing spike-time dependent
plastic inhibitory synapses similar to the network used by Vogels and
collaborators. Random, possibly overlapping patterns are iteratively stored
in the network as Hebbian assemblies. In order to measure their signal to
noise ratio, the stored patterns are recalled by the activation of a randomly
selected subset of their neurons. In general, during the recall of the stored
patterns, we observe a gradual decrease in the signal to noise ratio as
more and more patterns are stored in the network. This signifies an inverse
relation between the mean signal to noise ratio and the total number of
such patterns that were stored in the network. We are currently
investigating the performance of our balanced networks during and after
focal lesions and homoeostatic rewiring.
References
1. Vogels T, Sprekeler H, Zenke F, Clopath C, Gerstner W: Inhibitory plasticity
balances excitation and inhibition in sensory pathways and memory
networks. Science 2011, 334(6062):1569-1573.
2. Butz M, van Ooyen A: A simple rule for dendritic Spine and axonal
bouton formation can account for cortical reorganization after focal
retinal lesions. PLoS Comput Biol 2013, 9(10):e1003259.
3. Butz M, van Ooyen A: Homeostatic structural plasticity - a key to
neuronal network formation and repair. BMC Neuroscience 2014,
15(Suppl 1):P17.
4. Keck T, Mrsic-Flogel TD, Afonso MV, Eysel UT, Bonhoeffer T, Hbener M:
Massive restructuring of neuronal circuits during functional
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
reorganization of adult visual cortex. Nature Neuroscience 2008,
11(10):1162-1167.
5. Yamahachi H, Marik SA, McManus JN, Denk W, Gilbert CD: Rapid axonal
sprouting and pruning accompany functional reorganization in primary
visual cortex. Neuron 2009, 64(5):719-729.
6. Dayan P, Willshaw DJ: Optimising synaptic learning rules in linear
associative memories. Biological Cybernetics 1991, 65(4):253-265.
P236
Using transfer entropy to study synaptic integration in Purkinje cells
Kirsty Kidd*, Neil Davey, Daniel Polani, James M Bower, Volker Steuber
School of Computer Science, University of Hertfordshire, Hatfield, UK
E-mail: k.kidd2@herts.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P236
Information theory has been used in many ways to analyse the output of
neurons and how it relates to the input. Mutual information is often used
as a means to describe this relationship [1], but as it is symmetrical, it does
not necessarily describe how much information is transferred specifically
from input to output. An alternative method, transfer entropy [2],
introduces directionality to the analysis of this problem. Transfer entropy
allows one to study how the predictability of the output of the cell is
affected by varying time windows of preceding input spike trains.
In this work we are using two different extensions to transfer entropy that
attempt to limit bias. The first method of estimating transfer entropy we
are using is as described by Wibral et al [3], which places a restriction of a
single time step on the length of the delay considered on the destination
time series. This restriction is placed to avoid overestimating the
information transferred from input to output by ensuring that the
information provided by the history of the output has not been
underestimated. The second approach, put forward by Gourvitch and
Eggermont [4] creates a shuffled transfer entropy value from the average
transfer entropy, taken over many runs where the history of the input is
randomly shuffled. The shuffled value can then be used to normalise the
transfer entropy of the un-shuffled data.
We are currently investigating the use of transfer entropy to study the
output of an active Purkinje cell model [5,6] in response to a gamma-
distributed input, while varying the location of the input site. Active
channels in dendrites allow the effect of input to be location-independent,
but using transfer entropy enables us to highlight differences in delay of
information transfer based on the morphology of the cell.
References
1. London M, Schreibman A, Husser M, Larkum ME, Segev I: The Information
Efficacy of a Synapse. Nature Neuroscience 2002, 5(4):332-340.
2. Schreiber T: Measuring Information Transfer. American Physical Society
2000, 85(2):461-464.
3. Wibral M, Pampu N, Priesemann V, Siebenhhner F, Seiwert H, Lindner M,
Lizier JT, Vicente R: Measuring Information-Transfer Delays. PLoS One 2013,
8(2).
4. Gourvitch B, Eggermont JJ: Evaluating Information Transfer Between
Auditory Cortical Neurons. Journal of Neurophysiology 2007, 97:2533-2543.
5. De Schutter E, Bower JM: An Active Membrane Model of the Cerebellar
Purkinje Cell I. Simulation of Current Clamps in Slice. Journal of
Neurophysiology 1994, 71(1):375-400.
6. De Schutter E, Bower JM: An Active Membrane Model of the Cerebellar
Purkinje Cell II. Simulation of Synaptic Responses. Journal of
Neurophysiology 1994, 71(1):401-419.
P237
Animal control by spiking neural networks evolved with a genetic algorithm
Borys Wrbel1,2*, Ahmed Abdelmotaleb1,3, Neil Davey3, Volker Steuber3
1 1
Evolutionary Systems Group, Adam Mickiewicz University, Poznan, Poland;
2
Systems Modeling Group, IOPAN, Sopot, Poland; 3Biocomputation Research
Group, University of Hertfordshire, Hatfield, UK
E-mail: wrobel@evosys.org
BMC Neuroscience 2015, 16(Suppl 1):P237
The ability to search for resources is an example of a minimally cognitive
behaviora behavior shown by even the simplest animals, and that can
be explored using simple robots [1]. Even very simple networks (such as
Page 157 of 200
natural and artificial genetic or neural networks) allow for control of the
simplest search behaviors [2]. Moreover, this cognitive task can be made
more difficult [3], so it can be seen as a possible stepping step toward
advanced cognitive skills, both in biology and robotics [1].
In biology, the topology and synaptic weights of simple networks is
rather evolved than learned. Here we used an artificial life platform called
GReaNs that allows to use a genetic algorithm to evolve simple spiking
neural networks (SNNs) using a mixed bio-inspired paradigm - the way
the topology and weighs are encoded in artificial genomes is inspired by
genetic networks, but computational units in the network are modeled as
either leaky integrate and fire neurons with a fixed threshold or adaptive-
exponential integrate and fire neurons [3].
We evolved SNNs with GReaNs to control simple artificial robots (animats)
whose task was to search for targets in a 2-dimensional artificial
environment. The targets can be seen as food pellets from which food
diffuses, and is sensed by robots two smell/taste sensors, on the left and
right front. The robot also has two actuators which generate thrust on
the left and right back; when the thrust on one side is larger, the animat
moves in a circle; when the thrusts are the same, the robot goes forward
in a straight line.
We have designed three ways to present the strength of the sensed signal
to the network, and three ways to relate the thrust generated by the two
actuators to the activity of the corresponding two motor neurons in the
network. All the tested setups allowed us to evolve robots with correct
search behavior. Out of three setups for sensors, two can be seen as
biologically realistic. In one of them, easier to evolve, the sensory
information was preprocessed. Pre-processing consisted of calculating the
difference and the sum of the smell sensed by the sensors, using the
results as arguments of two sigmoid functions to obtain two values that
determined the percentages of activation of two populations of 100
primary sensory neurons, each connected with one synapse (with the
same weight) to secondary sensory neurons. In the second setup, a Hill
function was used to map the smell of two sensors as current injection to
two sensory neurons (in other words, here there was no pre-processing of
the difference between the smell strength on two sides of the robot). Out
of two setups for actuators we tested, again two were biologically realistic.
It proved easier to evolve a setup in which constant thrust was generated
in an actuator when the corresponding motor neurons spiked. In the less
evolvable approach, the thrust was determined by summing the number
of spikes of the corresponding motor neuron over a sliding temporal
window.
Acknowledgements: The work in BWs lab is supported by the Polish
National Science Center (project EvoSN, UMO-2013/08/M/ST6/00922). AA
was supported by the Foundation for Polish Science, co-financed by EU
Regional Development Fund (Innovative Economy Operational
Programme 2007-2013).
References
1. Wrbel B: Challenges for a-life approach to artificial cognition: in search
for hierarchy of cognitive systems. Artificial Life XIII: Proceedings of the
Thirteenth International Conference on the Simulation and Synthesis of Living
Systems Cambridge, MA: MIT Press 2012, 599-600.
2. Wrbel B, Joachimczak M: Using the Genetic Regulatory evolving Artificial
Networks (GReaNs) platform for signal processing, animat control, and
artificial multicellular development. Growing Adaptive Machines Berlin,
Heidelberg: Springer-Verlag 2014.
3. Braitenberg V: Vehicles: Experiments in Synthetic Psychology Cambridge, MA:
MIT Press 1984.
P238
Evolving small spiking neural networks to work as state machines for
temporal pattern recognition
Borys Wrbel1,2*, Ahmed Abdelmotaleb1,3, Neil Davey3, Volker Steuber3
Evolutionary Systems Group, Adam Mickiewicz University, Poznan, Poland;
2
Systems Modeling Group, IOPAN, Sopot, Poland; 3Biocomputation Research
Group, University of Hertfordshire, Hatfield, UK
E-mail: wrobel@evosys.org
BMC Neuroscience 2015, 16(Suppl 1):P238
The mechanisms that allow biological networks to recognize temporal
patterns of spikes that encode sensory inputs are unclear. Here we
extend our previous work [1], using an artificial life software platform,
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 158 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

GReaNs [2] to evolve spiking neural networks as state machines to

recognize temporal patterns of spikes.
GReaNs implements a genetic algorithm to obtain the topology of the
networks (and the weights of the synaptic connections), starting from a
population of networks of neurons connected randomly. The encoding of
the neural networks in the genome is inspired by the encoding of genetic
networks in biological genomes; neurons in GReaNs are modeled as either
leaky integrate and fire neurons with a fixed threshold (LIF) or adaptive-
exponential integrate and fire neurons [2]. The number of neurons in the
network is not limited in GReaNs, but here as previously [1] we limit the
size of the networks so that the analysis of the way the networks function
is simplified.
In the computational task we consider, the network has several input
neurons and one output neuron. A spike or burst received by an input
neuron corresponds to a certain symbol (for example, sound with a
specific frequency; a flash of light with a specific color). The output neuron
should be active only after the network receives a certain sequence of
symbols (a temporal pattern).
Our preliminary results with LIF networks with a fixed threshold networks
suggest that the presence of recurrent connections in the network allows
the interneurons to reach plateau subthreshold states that provide a
memory of what symbols have been received thus far. Here we will
investigate the robustness of this solution to noise in the network, and
then discuss the possibility to extend the paradigm to evolve spiking
networks to accept regular languages.
Acknowledgements: The work in BWs lab is supported by the Polish
National Science Center (project EvoSN, UMO-2013/08/M/ST6/00922). AA
was supported by the Foundation for Polish Science, co-financed by EU
Regional Development Fund (Innovative Economy Operational
Programme 2007-2013). Figure 1(abstract P239) A) Detectability of synaptic inputs of
References different strengths in a fully-defined input paradigm (data for N =
1. Abdelmotaleb A, Davey N, Schilstra M, Steuber V, Wrbel B: Evolving 4 Layer 2/3 pyramidal cells) for 1Hz, 5Hz, and 10Hz post-synaptic
spiking neural networks for temporal pattern recognition in the firing rates. Note that detection of inputs improves with higher firing
presence of noise. Artificial Life XIV: Proceedings of the Fourteenth rates and that inhibitory inputs require longer recording periods than
International Conference on the Simulation and Synthesis of Living Systems detection of excitatory inputs. Detection time for inputs of different
Cambridge, MA: MIT Press 2014, 966-972. amplitudes drops as approximately c/x2. B, left) Ground truth PSC
2. Wrbel B, Joachimczak M: Using the Genetic Regulatory evolving Artificial amplitude vs the amplitude of the cross-correlation (25ms window
Networks (GReaNs) platform for signal processing, animat control, and following pre-synaptic spikes) for 5Hz post-synaptic spiking. B, right)
artificial multicellular development. Growing Adaptive Machines Berlin, Ground truth PSC amplitude vs the model-based coupling coefficient.
Heidelberg: Springer-Verlag 2014. Note that model-based coefficients more accurately reconstruct the true
amplitudes. Dashed lines denote a linear fit, gray line denotes a 4th-
order polynomial fit to account for the nonlinear relationship between
P239 PSC amplitude and correlation coefficients
Identifying and tracking simulated synaptic inputs from neuronal firing:
insights from in vitro experiments
Maxim Volgushev1, Vladimir Ilin1, Ian H Stevenson1,2*
1 from spikes relate to simulated synaptic input? and 2) What are the
Department of Psychology, University of Connecticut, Storrs, CT, USA;
2
Department of Biomedical Engineering, University of Connecticut, Storrs, CT, limitations of connectivity inference?
USA Using likelihood-based models of functional connectivity, we find that
E-mail: ian.stevenson@uconn.edu individual current-based synaptic inputs are detectable over a broad range
BMC Neuroscience 2015, 16(Suppl 1):P239 of amplitudes and conditions. Detectability depends on input amplitude
and output firing rate, and excitatory inputs are detected more readily
Accurately describing synaptic interactions between neurons and how than inhibitory (Figure 1). Moreover, as we model increasing numbers of
interactions change over time are key challenges for systems neuroscience. presynaptic inputs, we are able to estimate connection strengths more
Although intracellular electrophysiology is a powerful tool for studying accurately and detect the presence of connections more quickly. We find
synaptic integration and plasticity, it is limited by the small number of that model-based methods outperform previous methods for detecting
neurons that can be recorded simultaneously in vitro and by the technical
synaptic connections (nonparametric tests on the cross-correlation), and
difficulty of intracellular recording in vivo. One way around these
difficulties may be to use large-scale extracellular recording of spike trains simulation results (adaptive exponential integrate-and-fire) suggest that
and apply statistical methods to model and infer functional connections the model results hold even when synaptic input is conductance-based
between neurons. These techniques have the potential to reveal large- rather than current-based. Together these results illustrate the possibilities
scale connectivity structure based on the spike timing alone. However, the and outline the limits of inferring synaptic input from spikes.
interpretation of functional connectivity is often approximate, since only a
small fraction of presynaptic inputs are typically observed.
Here we use in vitro current injection in layer 2/3 pyramidal neurons to
validate methods for inferring functional connectivity in a setting where P240
input to the neuron is controlled. In experiments with partially-defined Neural representation of a spatial odor memory in the honeybee
input we inject a single simulated input with known amplitude on a mushroom body
background of fluctuating noise. In a fully-defined input paradigm we then Martin P Nawrot1,2*, Tiziano DAlbis1, Randolf Menzel3, Martin Strube-Bloss4
1
control the synaptic weights and timing of many (1024) simulated Bernstein Center for Computational Neuroscience Berlin, Berlin, Germany;
2
presynaptic neurons. By analyzing the firing of neurons in response to Computational Systems Neuroscience, Department of Biology, University of
these artificial inputs, we ask 1) How does functional connectivity inferred Cologne, Cologne, Germany; 3Institute of Biology - Neurobiology, Freie
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Universitt Berlin, Berlin, Germany; 4Department of Behavioral Physiology &
Sociobiology, Biocenter, University of Wrzburg, Wrzburg, Germany
E-mail: martin.nawrot@fu-berlin.de
BMC Neuroscience 2015, 16(Suppl 1):P240
Insects make use of lateralized olfactory information from their left and
right antennae. Honeybees can learn to distinguish side-specific odor cues
in classical conditioning experiments, i.e. they associate a specific stimulus
combination of odor identity and spatial location (left or right) with the
reward [1]. This requires inter-hemispheric transfer of lateralized
information and a side-specific odor memory. Mushroom body (MB)
output neurons, the so-called extrinsic neurons (ENs), make inter-
hemispheric connections between the two MBs and are thus candidates
for the inter-hemispheric transfer of lateralized stimulus information. We
could show previously that ENs in the honeybee undergo plastic changes
in classical conditioning [2].
Here, we investigate neuronal plasticity in ENs of the honeybee in a sides-
specific learning paradigm. We performed multiple single-unit recordings
from ENs of one MB. Prior to conditioning (PRE) each bee was repeatedly
presented with two different odors on the two antennae separately.
During acquisition one of these odors was repeatedly presented to the
antenna contralateral to the recording side (CS+) and paired with a sugar
reward, while the other odor was presented without reward (differential
conditioning). Three hours after training (POST) we repeated the initial
protocol presenting each of the two odors repeatedly to each of the
antennae. In the behavioral test the bees distinguished the CS+ stimulus
configuration of odor and side from the other three stimulus combinations.
At the neuronal level we found clear and distinct odor representations in
the EN population before training (PRE) only when stimulated on the
antenna ipsilateral to the recording side. However, no odor responses
were measured in any of the ENs when stimulated at the contralateral
antenna. This picture changed drastically after training (POST). Now, the
rewarded stimulus combination (CS+) resulted in a strong population
response pattern. The population code for the CS+ configuration was
clearly distinct from all three other stimulus configurations. Quantification
of the temporal response latencies showed that the ENs encode an odor
approximately within 75ms for ipsilateral stimulation. Odor representation
was delayed by about 60ms for a contralateral stimulation. We discuss
two alternative explanations for this temporal delay.
We hypothesized previously that ENs at the MB output encode the
experience-dependent value of a particular stimulus [2-4]. Our results here
provide additional evidence for this hypothesis. A representation of the
rewarded stimulus combination (CS+) of a particular odor and its spatial
location (left or right) develops only due to reward conditioning. Before
learning, only ipsilateral odor information was represented.
Acknowledgements: Funding was received from the Federal Ministry of
Education and Research (BMBF) within the Bernstein Focus Neuronal Basis
of Learning (BFNL) through grant 01GQ0941 to M.N. and R.M.
References
1. Sandoz JC, Menzel R: Side-Specificity of Olfactory Learning in the
Honeybee: Generalization between Odors and Sides. Learning & Memory
2001, 8:286-294.
2. Strube-Bloss MF, Nawrot MP, Menzel R: Mushroom body output neurons
encode odor-reward associations. J Neurosci 2011, 31:3129-3140.
3. Menzel R: The insect mushroom body, an experience-dependent
recoding device. J Physiol Paris 2014, 108:84-95.
4. Kloppenburg P, Nawrot MP: Neural Coding: Sparse but On Time. Current
Biology 2014, 24:R957-R959.
P241
The effect of synchronized pauses on the coding strategies of
cerebellar nuclear neurons: a modeling study
Shyam Kumar1,2*, Benjamin Torben-Nielsen1, Erik De Schutter1,2
1
Computational Neuroscience Unit, Okinawa Institute of Science and
Technology, Onna-son, Okinawa, Japan; 2Department of Theoretical
Neurobiology and Neuroengineering, University of Antwerp, Wilrijk, Belgium
E-mail: shyam_u2@oist.jp
BMC Neuroscience 2015, 16(Suppl 1):P241
Purkinje cells (PC) of the cerebellar cortex have two distinct firing
signatures: simple and complex spikes. The simple spikes are due to the
Page 159 of 200
intrinsic mechanisms of the cell and synaptic inputs coming through the
parallel fibers and molecular layer interneurons [1-3]. The PCs also emit
complex spikes, which are due to strong excitation coming from the
axons of the inferior olive cells, the climbing fibers [2]. The simple spikes
are highly regular intertwined with short pauses [4] while complex spikes
occur sporadically and consist of burst of spikelets [5-7]. How can the
cerebellar nuclei neurons downstream of the PCs make sense of this
complex firing pattern?
In this study we analyze how PC synchrony in the context of pauses in
simple spikes affects different coding mechanisms of the neurons of the
cerebellar nuclei. To this end, we use a computational model of a
cerebellar nuclear neuron [8] and synthetic PC spike trains. The coding
mechanisms of cerebellar nuclear neuron can be broadly categorized as
rate [9,10] and time coding [11]. We define PC synchrony as synchronized
Purkinje neuron pauses with either pause beginning or pause ending
spikes precisely synchronized [4]. With varying amount of pause
synchrony for the above mentioned synchrony types, we analyzed its
effects on time locking and rate coding in a nuclear cell.
We find that both the amount and type of pause synchrony is encoded
as rate increases in the firing of the nuclear cell. Synchronized pauses
briefly release the nuclear neuron from inhibition giving rise to well-
timed spikes. Further, pauses synchronized with pause ending spikes
caused greater firing rate modulation in the nuclear cell while pause
beginning type synchrony enhanced the degree of timelocking. We also
analyze the effect of pause length and spike jitter on the timelocking
phenomenon of the nuclear neuron. We argue that these results lead to
better understanding of how PC pause synchrony is processed in its
target nuclear neuron.
References
1. Raman IM, Bean BP: Ionic Currents Underlying Spontaneous Action
Potentials in Isolated Cerebellar Purkinje Neurons. J Neurosci 1999,
19(5):1663-1674.
2. Palay SL, Chan-Palay V: Cerebellar Cortex 1974.
3. Gundappa-sulur G, Schutter EDE, Bower JM: Ascending Granule Cell Axon: An
Important Component of Cerebellar Cortical Circuitry 1999, 408:580-596.
4. Shin SL, De Schutter E: Dynamic synchronization of Purkinje cell simple
spikes. Journal of Neurophysiology 2006, 96(6):3485-3491.
5. Eccles JC, Llins R, Sasaki K: The excitatory synaptic action of climbing
fibres on the Purkinje cells of the cerebellum. J Physiol 1966,
182(2):268-296.
6. Llins R, Sugimori M: Electrophysiological properties of in vitro Purkinje
cell dendrites in mammalian cerebellar slices. J Physiol 1980, 305:197-213.
7. Shin SL, Rotter S, Aertsen A, De Schutter E: Stochastic description of
complex and simple spike firing in cerebellar Purkinje cells. The European
Journal of Neuroscience 2007, 25(3):785-794.
8. Steuber V, Schultheiss NW, Silver RA, De Schutter E, Jaeger D: Determinants
of synaptic integration and heterogeneity in rebound firing explored
with data-driven models of deep cerebellar nucleus cells. Journal of
Computational Neuroscience 2011, 30(3):633-658.
9. Armstrong BYDM, Rawson JA: Activity patterns of cerebellar cortical neurones
and climbing fibre afferents in the awake cat 1979, 289:425-448.
10. McDevitt CJ, Ebner TJ, Bloedel JR: Relationships between simultaneously
recorded Purkinje cells and nuclear neurons. Brain Research 1987,
425(1):1-13.
11. Person AL, Raman IM: Purkinje neuron synchrony elicits time-locked
spiking in the cerebellar nuclei. Nature 2012, 481(7382):502-505.
P242
Modeling of seizure transitions with ion concentration dynamics
Damiano Gentiletti1*, Marco De Curtis2, Vadym Gnatkovski2, Piotr Suffczynski1
1
Department of Experimental Physics, University of Warsaw, Warsaw, Poland;
2
Fondazione Istituto Neurologico Carlo Besta, Milan, Italy
E-mail: gentiletti.damiano@gmail.com
BMC Neuroscience 2015, 16(Suppl 1):P242
Traditionally it is considered that neuronal synchronization in epilepsy is
caused by a chain reaction of synaptic excitation. However, it has been
shown that synaptic transmission is not necessary for epileptiform
synchronization [1]. In order to investigate the respective roles of synaptic
and non-synaptic neuronal coupling in seizure transitions, we developed a
computational model of hippocampal network, involving extracellular
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
space, realistic dynamics of Na + , K + and Cl - ions, the glial uptake and
diffusion mechanism. We show that network behavior under synaptic
coupling conditions may be quite different from the neurons activities
when specific non-synaptic components are included. In particular, we
show that in the extended model, strong discharge of inhibitory
interneurons may result in long lasting accumulation of extracellular K+,
which sustains depolarization of principal cells and causes their
pathological discharges. This effect is not present in a reduced, purely
synaptic network. These results confirm the experimental hypothesis that
increase of inhibitory interneurons firing may lead to increased firing in the
pyramidal cells through accumulation of extracellular potassium [2]. The
model also shows that all potassium clearance mechanisms (glial uptake,
sodium-potassium pump, potassium diffusion) are critically important to
reproduce the experimental findings. This means that computational
modeling of seizure activity without ion dynamics may lead to unrealistic
results.
References
1. Jefferys JGR, Haas HL: Synchronized bursting of CA1 hippocampal
pyramidal cells in the absence of synaptic transmission. Nature 1982,
300:448-450.
2. De Curtis M, Gnatkovsky V: Reevaluating the mechanisms of focal
ictogenesis: the role of low-voltage fast activity. Epilepsia 2009,
50(12):2514-2525.
P243
Lateral Inhibition as the organizer of the bottom-up attentional
modulation in the primary visual cortex
Elbieta Gajewska-Dendek1*, Andrzej Wrbel2, Piotr Suffczynski1
1
Department of Biomedical Physics, Institute of Experimental Physics,
University of Warsaw, Warsaw, 02-093, Poland; 2Department of
Neurophysiology, Nencki Institute of Experimental Biology, Warsaw, 02-093,
Poland
E-mail: egdfuw@edu.pl
BMC Neuroscience 2015, 16(Suppl 1):P243
Lateral inhibition is known to occur at all levels of processing of visual
information: from retinal ganglion cells, through principal cells in LGN to
primary cortical areas [1]. The role of this connectivity module is to
contrast lateral variation within the simultaneously processed upstream
information.
Our work is based on local field potential (LFP) multichannel signal data [2]
recorded from the cats visual cortices activated by attentional tasks:
voluntary (top-down) and stimulus driven (bottom-up). These experiments
showed that enhanced beta signals (16-24 Hz) serve as a carrier for
distributing attentional activation across the visual system. By measuring
the correlation strengths between spontaneous LFP signals in different
cortical recording sites it was also found that beta activation maps were
organized in a mosaic-like pattern within the visual cortex during bottom-
up task, whereas top-down task resulted in spatially homogeneous cortical
beta excitation.
It was hypothesized that the reason for these different cortical activation
patterns was due to lateral inhibition raised by salient visual stimulus
during bottom-up task [2]. This work aims to verify a role played by lateral
inhibition in producing specific spatial activation patterns found in top-
down and bottom-up experimental paradigms by means of computational
modelling.
We have constructed a model of neuronal network comprised of single
compartment excitatory and inhibitory cells activated according to
extended Hodgkin-Huxley dynamics. The network consists of 16 domains
representing cortical patches [3], where the neuronal connections keep
the anatomical weights, proportions and local structure according to
experimental data: neurons are more likely to form synapses with
neighboring cells within one domain. Additionally, the domains are more
likely to be interconnected with adjacent than with distant ones. LFPs are
modelled as a sum of synaptic currents taking into account their
distances from a point of measurement.
The modelled neurons receive two kinds of Poisson inputs, which
represent the feed-forward input from the thalamus and feed-back
activation from higher order cortical regions. The reinforcement of the
latter causes widespread increase in beta band power of LFP and
represents diffuse, top-down attentional modulating signal. The input
Page 160 of 200
from the thalamus arriving to one or more domains represents bottom-
up attentional activation of excitatory and inhibitory neurons evoked by
salient visual stimulus. The cortical domains activated by this salient input
exert additional lateral inhibition on neighboring domains.
Our model shows that the spatial heterogeneity of cortical activation
resulting in local decrease of correlation in bottom-up task and
homogeneity in top-down task, can be mediated by feed-forward lateral
inhibitory influences at the cortical level. The higher cross-correlation
variance observed experimentally during bottom-up than in top-down
attentional paradigm is also observed in the model.
This modelling study, confirmed that mechanism of lateral inhibition may
be sufficient in explaining spatial differences in organization of beta
activities in visual cortical areas observed during bottom-up and top-
down attentional paradigms.
The network was modelled by means of PyNEST package.
References
1. Isaacson JS, Scanziani M: How Inhibition Shapes Cortical Activity. Neuron
2011, 72:231-243.
2. Wrbel A: Attentional Activation in Corticothalamic Loops of the Visual
System. The New Visual Neurosciences (Werner JS, Chalupa LM, eds)
Cambridge, MA: MIT Press 2014, 339-349.
3. Douglas RJ, Martin KA: Neuronal circuits of the neocortex. Annual Review
of Neuroscience 2004, 27:419-451.
P244
Computational interactions between decision and emotion
Nicoladie D Tam
Department of Biological Sciences, University of North Texas, Denton, TX
76203, USA
E-mail: nicoladie.tam@unt.edu
BMC Neuroscience 2015, 16(Suppl 1):P244
Introduction: Decision is often influenced by emotions, such that the
decision can often be biased by the emotional influences, such as
happiness, sadness, fear and anger. Previous experimental studies in
human subjects have shown that decisions are often related to emotional
levels [1,2]. It is also often assumed that decision is influenced by
emotion, but there is evidence that decision can also influence emotions.
Thus, the interrelationship between decision and emotion requires in-
depth re-examination to determine the interactions between decision
and emotion.
Methods:: A computational model of decision-making relative to emotion
is derived based on the experimental evidence of decision in relation to
emotion in human subjects. Using the classical behavioral economic
experimental Ultimatum Game paradigm [3] that elicits the
interrelationship between decision and emotion in human subjects [4-7],
a computational model of decision is derived based on the shifting of the
threshold in the stimulus-response function of emotional responses.
Results:: Using the quantitative analysis, the emotional stimulus-response
function was derived based on the disparity between the desirable
predicted outcome and the actual outcome in the real world for happy
[4], sad [6], angry [5], jealous [7] emotions and for fairness perception [8],
using the optimizations in survival functions [9,10]. The relationships
between decision and emotions were also established for happy [1]
emotion, and for fairness [11] experimentally, and the interrelationship
between decision and fairness was derived computationally [12].
Extending the result to emotions, let e be the vector representing the
emotional intensity, d be the vector representing the disparity between
the predicted outcome and actual outcome, then the emotional stimulus-
response function is given by: e = kf(d) + b where f() represents a
nonlinear function, and the coefficient k is the emotional sensitivity and b
is the emotional baseline level. Let x be the vector representing the
decision (where x = 1 represents a yes decision, and x = -1 represents a
no decision), then the decision threshold can be given by:

1, if kf (d) + b
x= where is the decision threshold,
1, otherwise
representing the dependence of decision on emotion.
Discussion:: The interrelationship between decision and emotion can be
derived based on the threshold crossing of the emotional intensity level,
in which an emotional bias in either the emotional baseline or the
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
emotional sensitivity can cause a change in decision. The decision is
based on the threshold crossing of the emotional stimulus-response
function, such that it is a continuum in altering the decision-making
process by the shifting of emotional bias in either emotional baseline or
sensitivity.
References
1. Tam ND: Quantification of happy emotion: Dependence on decisions.
Psychol Behav Sci 2014, 3(2):68-74.
2. Tam ND: Rational decision-making process choosing fairness over
monetary gain as decision criteria. Psychol Behav Sci 2014, 3(6-1):16-23.
3. von Neumann J, Morgenstern O, Rubinstein A: Theory of games and
economic behavior Princeton, NJ: Princeton University Press 1953.
4. Tam ND: Quantification of happy emotion: Proportionality relationship to
gain/loss. Psychol Behav Sci 2014, 3(2):60-67.
5. Tam DN: Computation in emotional processing: quantitative
confirmation of proportionality hypothesis for angry unhappy emotional
intensity to perceived loss. Cogn Comput 2011, 3(2):394-415.
6. Tam ND: Quantitative assessment of sad emotion. Psychol Behav Sci 2014,
4(2):36-43.
7. Tam ND, Smith KM: Cognitive computation of jealous emotion.
Psychology and Behavioral Sciences 2014, 3(6-1):1-7.
8. Tam ND: Quantification of fairness perception by including other-
regarding concerns using a relativistic fairness-equity model. Adv in Soc
Sci Research J 2014, 1(4):159-169.
9. Tam D: EMOTION-I model: A biologically-based theoretical framework for
deriving emotional context of sensation in autonomous control systems.
The Open Cybernetics and Systemics Journal 2007, 1:28-46.
10. Tam D: EMOTION-II model: A theoretical framework for happy emotion
as a self-assessment measure indicating the degree-of-fit (congruency)
between the expectancy in subjective and objective realities in
autonomous control systems. The Open Cybernetics and Systemics Journal
2007, 1:47-60.
11. Tam ND: Quantification of fairness bias in relation to decisions using a
relativistic fairness-equity model. Adv in Soc Sci Research J 2014, 1(4):169-178.
12. Tam ND: A decision-making phase-space model for fairness assessment.
Psychol Behav Sci 2014, 3(6-1):8-15.
P245
Differential temporal activation of oxy- and deoxy-hemodynamic signals
in optical imaging using functional near-infrared spectroscopy (fNIRS)
Nicoladie D Tam1*, George Zouridakis2
1 1
Department of Biological Sciences, University of North Texas, Denton, TX
76203, USA; 2Departments of Engineering Technology, Computer Science,
and Electrical and Computer Engineering, University of Houston, Houston,
TX, 77204, USA
E-mail: nicoladie@tamunt.edu
BMC Neuroscience 2015, 16(Suppl 1):P245
Background: Optical imaging of the brain based on near-infrared
spectroscopy (NIRS) can provide real-time measurements of the
hemodynamic signals that represent metabolic demands of the underlying
neural tissues. Functional imaging based on NIRS (fNIRS) can detect both
oxy-hemoglobin (oxy-Hb) and deoxy-hemoglobin (deoxy-Hb) levels related
to neural metabolic activity, whereas BOLD fMRI (blood-oxygen-level
dependent functional magnetic resonance imaging) can only detect
signals related to deoxy-Hb. Thus, during task execution, only fNIRS can
determine the differential temporal activation/deactivation of oxy-Hb and
deoxy-Hb hemodynamic signals as the blood-oxygen demand changes.
We have previously shown that as metabolic demand increases, temporal
changes in oxy-Hb and deoxy-Hb levels can be temporally decoupled (i.e.,
oxy-Hb level can decrease while deoxy-Hb level increases) rather than
being coupled, in which case both would increase or decrease
simultaneously [1-5]. In order to account for the observed differential
temporal decoupling of oxy-Hb and deoxy-Hb levels, we hypothesize that
as oxygen demand increases, the delivery of blood oxygen cannot keep up
with the demand of the neural tissues, resulting in decreased oxy-Hb and
increased deoxy-Hb levels. This study provides experimental evidence that
validates the above hypothesis.
Methods: Human subjects were recruited to execute voluntary arm
movements in orthogonal directions to exert different oxygen demands
onto the motor cortex. The hemodynamic activities were recorded from
Page 161 of 200
the motor cortex using fNIRS, while the subjects executed predefined arm
movements. The oxy-Hb and deoxy-Hb levels were computed from the
NIRS optical signals using the modified Beer-Lambert law [6].
Results: Figure 1 shows the differential changes of oxy- and deoxy-Hb
hemodynamic signals over time during right and left movement
directions. These data demonstrate that the oxy- and deoxy-Hb
hemodynamic signals can change differentially rather than being coupled
in time. The differential changes in oxy- and deoxy-Hb levels can be
accounted by an oxygen demand exceeding the oxygen delivery in the
blood vessels.
Conclusions: The metabolic demands of the neural tissues are not
necessarily correlated with either oxy- or deoxy-Hb alone, but they are
correlated with the combination of both oxy- and deoxy-Hb. A decrease in
oxy-Hb level does not necessarily imply that oxygen demand decreases.
Rather, such a decrease in oxy-Hb level can be due to the rate of oxygen
demand by the neural tissues that exceeds the oxygen delivery capacity of
the blood vessels.
References
1. Tam ND, Zouridakis G: Optical imaging of motor cortical hemodynamic
response to directional arm movements using near-infrared
spectroscopy. International Journal of Biological Engineering 2013, 3(2):11-17.
2. Tam ND, Zouridakis G: Temporal decoupling of oxy- and deoxy-
hemoglobin hemodynamic responses detected by functional near-
infrared spectroscopy (fNIRS). Journal of Biomedical Engineering and
Medical Imaging 2014, 1(2):18-28.
3. Tam ND, Zouridakis G: Decoding movement direction from motor cortex
recordings using near-infrared spectroscopy. Infrared Spectroscopy: Theory,
Developments and Applications Hauppauge, NY: Nova Science Publishers, Inc
2014.
4. Tam ND, Zouridakis G: Decoding of movement direction using optical
imaging of motor cortex. BMC Neuroscience 2013, 2013:P380, 8 July 2013.
5. Tam ND, Zouridakis G: Optical imaging of motor cortical activation using
functional near-infrared spectroscopy. BMC Neuroscience 2012, 2012:P27,
16 July 2012.
6. Cope M, Delpy DT, Reynolds EO, Wray S, Wyatt J, van der Zee P: Methods
of quantitating cerebral near infrared spectroscopy data. Advances in
experimental medicine and biology 1988, 222:183-189.
P246
Mapping the smoking addiction using dynamic causal modelling at rest
Rongxiang Tang1, Adeel Razi2, Yi-Yuan Tang3*
Department of Psychology, The University of Texas at Austin, Austin, TX
78712, USA; 2The Wellcome Trust Centre for Neuroimaging, University
College London, London WC1N 3BG, UK; 3Department of Psychological
Sciences, Texas Tech University, Lubbock, TX 79409, USA
E-mail: yiyuan.tang@ttu.edu
BMC Neuroscience 2015, 16(Suppl 1):P246
Tobacco use is the leading preventable cause of death. Previous research
show that brain areas including medial prefrontal cortex (mPFC), posterior
cingulate cortex (PCC) and others involved in smoking addiction [1].
However it remains unclear which brain regions play a crucial role in
smoking addiction and the relationship among these regions. Since
functional connectivity does not support inferences about causal brain
connections, the changes in information flow in these distributed systems
involved in smoking remain largely unknown. Here we apply a dynamic
causal modeling (DCM) in resting state fMRI [2] to demonstrate the causal
relationships among the core regions in smoking addiction.
Healthy college students were recruited through campus advertisements.
Among those who responded, there were 14 cigarette smokers and 14
nonsmokers. All data were collected using a 3-Telsa Siemens Skyra
scanner. Functional data were processed using the Data Processing
Assistant for Resting-State fMRI, which is based on SPM (www.fil.ion.ucl.
ac.uk/spm) and Resting-State fMRI Data Analysis Toolkit. Based on
literature, we specified four regions of interest (ROIs) within the default
mode network (DMN), which are medial prefrontal cortex (mPFC),
posterior cingulate cortex (PCC), and bilateral inferior parietal lobule ( Left
IPL and Right IPL), same regions and coordinates as in previous studies
that use sDCM analysis for resting state [2,3]. Based on SPM12, we
estimated and specified the DCM for each subject and later compared
the differences of effective connectivity between two groups by using
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P245) Optical hemodynamic signals recorded from the motor cortex, showing different oxygen demands. (A) Rightward arm
movement. (B) Leftward arm movement. [Oxy-Hb (in red) and deoxy-Hb (in blue)]
two-sample test. Our results suggest the different causal relationships
between nonsmokers and smokers. Specifically, there was a lower
strength of excitatory input from PCC to mPFC for smokers than
nonsmokers and a lower strength of excitatory input from RIPL to mPFC
for smokers than nonsmokers (all p<0.05).
Conclusions: These data indicate the usage of DCM on resting-state fMRI
data can differentiate the causal brain connections between two groups,
and provide insight into the brain mechanisms underlying smoking
addiction - the abnormalities of causal connectivity associated with
attention and self-control networks in the brain. Our results may also
suggest the brain based prevention and intervention should consider the
amelioration of the mPFC-PCC circuits.
Acknowledgements: This work was supported by the Office of Naval
Research.
References
1. Goldstein RZ, Volkow ND: Dysfunction of the prefrontal cortex in
addiction: Neuroimaging findings and clinical implications. Nat Rev
Neurosci 2011, 12:652-669.
2. Razi A, Kahan J, Rees G, Friston KJ: Construct validation of a DCM for
resting state fMRI. Neuroimage 2015, 106:1-14.
3. Di X, Biswal BB: Identifying the default mode network structure using
dynamic causal modeling on resting-state functional magnetic
resonance imaging. Neuroimage 2014, 86:53-9.
P247
Brief mindfulness training alters causal brain connections in mTBI
Rongxiang Tang1, Yi-Yuan Tang2*
1
Department of Psychology, The University of Texas at Austin, Austin, TX
78712, USA; 2Department of Psychological Sciences, Texas Tech University,
Lubbock, TX 79409, USA
E-mail: yiyuan.tang@ttu.edu
BMC Neuroscience 2015, 16(Suppl 1):P247
Traumatic brain injury (TBI) is a significant cause of disability in the United
States and the mild TBI (mTBI) is the most prevalent. Previous research
indicates the positive effect of mindfulness training on symptoms of
chronic mTBI such as cognitive functioning and emotion [1]. However it
remains unclear which brain regions play a crucial role in mTBI recovery
Page 162 of 200
and the causal relationship among these regions. Here we apply a
dynamic causal modeling (DCM) in resting state fMRI [2] to demonstrate
the causal relationships among the core regions involved in mTBI.
Fourteen veteran students with mTBI were recruited through campus
advertisements. We used 2 weeks of integrative body-mind training as
mindfulness intervention, previously reported in our series of
randomized studies [3,4]. All data were collected using a 3-Telsa
Siemens Skyra scanner. Functional data were processed using the Data
Processing Assistant for Resting-State fMRI, which is based on SPM and
Resting-State fMRI Data Analysis Toolkit. Based on literature, we
specified four regions of interest within default mode network (DMN) -
medial prefrontal cortex (mPFC), posterior cingulate cortex (PCC), and
bilateral inferior parietal lobule (Left IPL and Right IPL), same regions
and coordinates as in previous sDCM studies [2,5]. Based on SPM12, we
estimated and specified the DCM for each subject and later compared
the differences of effective connectivity before and after mindfulness.
Our results suggest the different causal relationships following
mindfulness training. Specifically, after training there was a significant
decrease in the strength of excitatory input from mPFC to PCC, and a
significant increase in the strength of inhibitory input from mPFC to
LIPL (all p<0.05).
Conclusions: Resting DCM can differentiate the causal brain connections
before and after mindfulness training, and provide insight into the brain
mechanisms of altered DMN dynamics underlying mTBI recovery,
suggesting the changes in information flow in these distributed systems
involved in mTBI intervention.
Acknowledgements: This work was supported by the Office of Naval
Research.
References
1. Azulay J, Smart CM, Mott T, Cicerone KD: A pilot study examining the
effect of mindfulness-based stress reduction on symptoms of chronic
mild traumatic brain injury/postconcussive syndrome. J Head Trauma
Rehabil 2013, 28:323-31.
2. Razi A, Kahan J, Rees G, Friston KJ: Construct validation of a DCM for
resting state fMRI. Neuroimage 2015, 106:1-14.
3. Tang YY, et al: Short-term meditation training improves attention and
self-regulation. Proceedings of the National Academy of Sciences, USA 2007,
104:17152-17156.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
4. Tang YY, et al: Central and autonomic nervous system interaction is
altered by short term meditation. Proceedings of the National Academy of
Sciences, USA 2009, 106:8865-70.
5. Di X, Biswal BB: Identifying the default mode network structure using
dynamic causal modeling on resting-state functional magnetic
resonance imaging. Neuroimage 2014, 86:53-9.
P249
Surround suppression and normalization in a model of coupled
balanced cortical networks with short-term synaptic plasticity
Sara Konrad*, Tatjana Tchumatchenko
Theory of Neural Dynamics, Max-Planck Institute for Brain Research, Frankfurt,
60438, Germany
E-mail: Sara.Konrad@brain.mpg.de
BMC Neuroscience 2015, 16(Suppl 1):P249
Introduction: The input currents to neurons in the cortex are under many
circumstances highly variable with a mean intensity considerably below
the firing threshold. As a consequence, the spiking activity of cortical
neurons is strongly fluctuating such that the coefficient of variation of the
inter-spike interval distribution of individual neurons is approximately
equal to 1, implying almost Poisson-like spiking. In addition to this
characteristic activity, the connectivity of excitatory and inhibitory neurons
in cortex is sparse and irregular. It has been shown in models that super-
threshold excitatory and strong inhibitory input currents, which nearly
cancel for individual neurons, can lead to this irregular spiking activity [1].
However, balanced networks of excitatory and inhibitory neurons are
characterized by a strictly linear relation between stimulus strength and
network firing rate, making it hard to perform more complex
computational tasks like the generation of receptive fields, multiple stable
activity states or normalization, which for example has been measured in
visual cortex (eg. [2,3]).
Synapses displaying activity dependent short-term plasticity (STP) have
been previously reported to give rise to a non-linear network response
with potentially multiple stable states for a given stimulus [4].
In this study, we analyze analytically and numerically the computational
properties of two interconnected balanced networks, receiving independent
stationary stimuli. This situation can be viewed as a simple instantiation of
Figure 1(abstract P249) A. Schematic illustration of the network topology. E1 and I1 represent excitatory and inhibitory neuronal populations of
network 1, receiving input from an external population X1. Same for network 2. Coupling of neuronal populations is indicated by the arrows, synaptic
strengths by the Jyz. B. Schematic illustration of the resulting activity in the network for constant stimulus X1 and varying stimulus X2. In this model we
observe the activity of network 1 being suppressed when activity in network 2 increases. Colors of lines represent the population corresponding to panel A
Page 163 of 200
two interconnected cortical columns. For an illustration of the network
topology, see Figure 1A. We demonstrate that these stimuli are normalized
by the system and that increasing the stimulus to one network, suppresses
the activity of the neighboring network (see Figure 1B). Thereby,
normalization and suppression are linear in stimulus strength when STP is
disabled and becomes non-linear with activity dependent synapses.
References
1. van Vreeswijk C, Sompolinsky H: Chaotic Balanced State in a Model of
Cortical Circuits. Neural Comput 1998, 10(6):1321-1371.
2. Carandini M, Heeger DJ, Movshon JA: Linearity and normalization in
simple cells of the macaque primary visual cortex. J Neurosci 1997,
17(21):8621-8644.
3. Levitt JB, Kiper DC, Movshon JA: Receptive fields and functional
architecture of macaque V2. J Neurophysiol 1994, 71(6):2517-2542.
4. Mongillo G, Hansel D, van Vreeswijk C: Bistability and spatiotemporal
irregularity in neuronal networks with nonlinear synaptic transmission.
Phys Rev Lett 2012, 108(15):158101.
P250
Effect of power-law ionic conductances in the Hodgkin and Huxley
model
Fidel Santamaria
Wodimu TekaUTSA Neurosciences Institute, University of Texas at San
Antonio, San Antonio, TX 78249, USA
E-mail: fidel.satamaria@utsa.edu
BMC Neuroscience 2015, 16(Suppl 1):P250
An increasing number of results show that the voltage and spiking
activity of a neuron follows scale free adaptation. Under such conditions,
the voltage, or firing rate, of a neuron cannot be characterized by a
unique time constant, instead these processes are characterized by
power-laws. Power-law behaviors suggest that the components
responsible for the voltage are strongly interacting across temporal
scales, slow processes affect the rates of fast processes and vice versa.
We recently introduced the fractional leaky integrate-and-fire model,
which we have used to replicate the firing rate activity of adapting
cortical neurons [1]. Our results have shown that spike rate adaptation
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
can be modeled by a fractional derivative of low order. Thus suggesting,
a strong interaction across conductances in cortical neurons.
this work we decided to study the effects of power-law behavior in a
biophysical model of spiking activity, the Hodgkin-Huxley model. The
classical Hodgkin-Huxley model is described by.
dV
N of how to deal with this inverse non-linear problem. In this work, we aim
C = gm (V Erest ) + gi (V Ei ) + I. at giving a first proof of concept to address the estimation of synaptic
dt i=1
where C is the capacitance; V, voltage; gm, the membrane conductance;
E rest , the resting potential; g i ={g Na , g K } the sodium or potassium
conductances with corresponding reversal potentials (Ei={ ENa, EK }); and I,
the input curret. The conductances are gK = gK n4 and
gNa = gNa m3 h , with gK and gNa the maximum conductances. The
gating variables n, m, and h are defined by the general equation.
dx
= x (V, t) (1 x) x (V, t)x References
dt
where x={n, m, h}. To implement power-law behavior in either gating
variable we substitute the classical derivative by the fractional derivative
of order h using the Caputo definition.
d f 1 t f (n) (t)
= du
dt  (n ) 0 (t u)+1n 3. Rudolph M, Piwkowska Z, Badoual M, Bal T, Destexhe A: A method to
where is the Gamma function, and n is the nearest integer larger than
h. In our case, n = 1 because this models how the interaction of previous
activity slows down the activation of the gating variables. The fractional
derivative value is the result of integrating the activity of the function
over all past activities weighted by a function that follows a power-law.
The weighted values are called the memory trace. Thus, the fractional
derivative provides information over all past activity, in contrast with the
classical derivative that only takes into account the value of the function
in the immediate previous time point. We have recently developed
efficient ways to computationally solve these equations [2].
Our results show the emergence of a wide range of spiking behaviors
(bursting, spiking and sub-threshold oscillations) in response to constant
stimulation as a function of the fractional order in the different activation/
inactivation variables. In the case of n, the neuron shows reduction if spiking
response and emergence of sub-threshold oscillations. While fractional h
results in bursting activity. This emergent richness in spiking activity, while
only modeling two conductances, allows study of the overall effects of
power-law behavior in neuronal activity. At the biophysical level, our results
suggest that voltage dependent ion channels that deviate from a classical
Markov process could increase the amount of spiking patterns in single cells,
thus increasing their information content capacity.
Acknowledgements: NSF DBI 1451032 (BRAIN-EAGER) and NSF EF
1137897.
References
1. Teka W, Marinov TM, Santamaria F: Neuronal Spike Timing Adaptation
Described with a Fractional Leaky Integrate-and-Fire Model. PLoS Comput
Biol 2014, 10(3):e1003526.
2. Marinov T, Ramirez N, Santamaria F: Fractional integration toolbox. fcaa
2013, 16(3):670-681.
P251
Estimation of the synaptic conductance in a McKean-model neuron
Antoni Guillamon1, Rafel Prohens2, Antonio E Teruel2, Catalina Vich2*
1
Dept. of Applied Mathematics I, EPSEB, Universitat Politcnica de Catalunya,
08028 Barcelona, Spain; 2Dept. of Mathematics and Computer Science,
Universitat de les Illes Balears, 07122, Palma, Spain
BMC Neuroscience 2015, 16(Suppl 1):P251
Estimating the synaptic conductances impinging on a single neuron
directly from its membrane potential is one of the open problems to be
solved in order to understand the flow of information in the brain.
Despite the existence of some computational strategies that give
Page 164 of 200
circumstantial solutions ([1-3] for instance), they all present the
inconvenience that the estimation can only be done in subthreshold
activity regimes. The main constraint to provide strategies for the
oscillatory regimes is related to the nonlinearity of the input-output curve
and the difficulty to compute it. In experimental studies it is hard to
obtain these strategies and, moreover, there are no theoretical indications
conductances when the neuron is spiking. For this purpose, we use a
simplified model of neuronal activity, namely a piecewise linear version of
the Fitzhugh-Nagumo model, the McKean model ([4], among others),
which allows an exact knowledge of the nonlinear f-I curve by means of
standard techniques of non-smooth dynamical systems. As a first step,
we are able to infer a steady synaptic conductance from the cells
oscillatory activity. As shown in Figure 1, the model shows the relative
errors of the conductances of order C, where C is the membrane
capacitance (C<<1), notably improving the errors obtained using filtering
techniques on the membrane potential plus linear estimations, see
numerical tests performed in [5].
1. Bdard C, Bhuret S, Deleuze C, Bal T, Destexhe A: Oversampling method to
extract excitatory and inhibitory conductances from single-trial membrane
potential recordings. Journal of Neuroscience Methods 2012, 210:3-14.
2. Lankarany M, Zhu W-P, Swamy S, Toyoizumi T: Inferring trial-to-trial
excitatory and inhibitory synaptic inputs from membrane potential
using Gaussian mixture Kalman filtering. Frontiers in Computational
Neuroscience 2013, 7(109).
estimate synaptic conductances from membrane potential fluctuations.
Journal of Neurophysiology 2004, 91(6):2884-2896.
4. Coombes S: Neuronal networks with gap junctions: A study of piecewise
linear planar neuron models. SIAM Journal of Applied Dynamical Systems
2008, 7(3):1101-1129.
5. Guillamon A, McLaughlin DW, Rinzel J: Estimation of synaptic
conductances. Journal of Physiology-Paris 2006, 100(1-3):31-42.
P252
Induction and consolidation of calcium-based homo- and
heterosynaptic potentiation and depression
Yinyun Li, Tomas Kulvicius, Christian Tetzlaff*
III Institute of Physics, Department of Computational Neuroscience, Georg-
August-University Gttingen, Bernstein Center for Computational
Neuroscience, Gttingen, 37077, Germany
E-mail: tetzlaff@phys.uni-goettingen.de
BMC Neuroscience 2015, 16(Suppl 1):P252
Synaptic plasticity serves as the physiological foundation for learning and
memory [1]. While homosynaptic plasticity is associative learning or
Hebbian-type plasticity, heterosynaptic plasticity reflects the synaptic
change without direct stimulation, i.e. non-associative plasticity [2]. However,
heterosynaptic plasticity is an important mechanism preventing run-away
synaptic dynamics and offers a potential mechanism to understand memory
allocation [2,3]. Experimental results show that the induction of
heterosynaptic plasticity as well as homosynaptic plasticity depends on the
postsynaptic calcium concentration [4]. We propose that heterosynaptic
plasticity can be induced by the postsynaptic calcium dynamics which can
be triggered by the back propagation of action potentials.
However, homosynaptic plasticity has an early-phase (< 3 hours) and a
late-phase state (> 8 hours) [1]. Experiments show that an early-phase
synaptic change can be transferred to a late-phase by the mechanisms of
synaptic tagging and consolidation (STC) [5,6]: (i) the changed synapse
get tagged and (ii) a strong activation enables in the postsynaptic neuron
the synthesis of plasticity-related proteins (PRP) which are transmitted
back to the tagged synapse[5,6]. We propose that the same STC
mechanism consolidating homosynaptic changes are also able to
consolidate heterosynaptic changes.
We combine a history spiking-dependent neuron [7] with calcium-based
synaptic plasticity rule [8] and synaptic consolidation mechanism [9] to
understand: (i) the mechanisms of inducing heterosynaptic plasticity by
which the inactive synapse can change its weight through the
postsynaptic calcium level triggered by the back propagation of the
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P251) Goodness of fit of the synaptic conductance parameter. Panel A represents the relative error versus the applied current for
a fixed value of C = 10-4. Red points represent the values of I1 (left points) and I2 (right points) for each gsyn. Panel B represents the relative error versus
the membrane capacitance for a fixed value of I=I1+10-3. In both panels, the different color traces correspond to different values of gsyn equally spaced
from 0.1 to 0.3. The rest of parameters are fixed as a=0.25, v0=0, w0=0, g=0.5, vsyn=0.25+a /2
shared neuron; and (ii) of the consolidation of heterosynaptic changes
based on the synaptic tagging and consolidation principle. For instance, a
strong stimulus transmitted by a group of synapses induces and
consolidates by the postsynaptic neuron heterosynaptic changes at other,
unrelated synapses. Our study provides a further step of understanding
how several mechanisms interact with each other to enable the formation
of computational important long-term changes or memories.
Acknowledgements: This research is funded by from the European
Communities Seventh Framework Program FP7/2007-as well as from the
Germany Ministry of Science Grant to the Gttingen Bernstein Center for
Computational Neuroscience.
References
1. Abraham WC: How long will long term potentiation last? Phil. Trans. R.
Soc. Lond. B 2003, 358:735-744.
2. Chistiakova M, Bannon NM, Bazhenov M, Volgushev M : Heterosynaptic
plasticity: multiple mechansims and muliple roles. Neuroscientist 2014,
20(5):483-498.
3. Rogerson T, Cai DJ, Frank A, Sano Y, Shobe J, et al: Synaptic tagging
during memory allocation. Nat Rev Neurosci 2014, 15:157-169.
4. Malenka RC, Kauer JA, Zucker RS, Nicoll RA: Postsynaptic calcium is sufficient for
potentiation of hippocampal synaptic transmission. Science 1988, 242:81-84.
5. Frey U, Morris R GM: Synaptic tagging and long-term potentiation. Nature
1997, 385:533-536.
6. Sajikumar S, Navakkode S, Frey JU: Identification of compartment-and
Process-Specific Molecules Required for Synaptic Tagging during Long-
Term Potentiation and Long-Term Depression in Hippocampal CA1.
J Neurosci 2007, 27(19):5068-5080.
7. Yamauchi S, Kim H and Shinomoto S: Elemental spiking neuron model for
reproducing diverse firing patterns and predicting precise firing times.
Front in Comput neurosci 2011, 5(42):1-15.
8. Graupner M, Brunel N: Calcium-based plasticity model explains sensitivity
of synaptic changes to spike pattern, rate and dendritic location. PNAS
2012, 109(10):3991-3996.
9. Clopath C, Ziegler L, Vasilaki E, Buesing L, Gerstner W: Tag-Trigger-
Consolidation: A Model of Early and Late Long-Term-Potentiation and
Depression. PLoS CB 2008, 4(12):e1000248.
P253
Interaction between memories in an abstract mathematical model
based on the Hebbian cell assembly hypothesis
Juliane Herpich1*, Florentin Wrgtter1,2, Christian Tetzlaff1,2
1
Third Physical Institute - Biophysics, Georg-August-University, Gttingen,
Germany; 2Bernstein Center for Computational Neuroscience, Gttingen,
Germany
E-mail: juliane.herpich@phys.uni-goettingen.de
BMC Neuroscience 2015, 16(Suppl 1):P253
Page 165 of 200
Learning and memory are essential properties of adaptive neural circuits.
Thereby, the general hypothesis (Synaptic-Plasticity-and-Memory
hypothesis; [1,2]) is that the adaptive process of synaptic plasticity induces
changes at the synapses connecting the neurons. These changes lead to
the formation of strongly interconnected subgroups of neurons, so-called
cell assemblies [3]. It has been suggested that such cell assemblies
represent the learned memory items. However, as known from everyday
life, after learning, humans and animals show the remarkable ability to
connect, generalize, and discriminate old and new memories. How these
memory interactions are realized on a neuronal level based on the idea of
cell assemblies is still unknown.
In this work, we use a network model dependent on the interaction
between synaptic plasticity and synaptic scaling [4,5]. Amongst others, this
interaction yields the formation of cell assemblies showing dynamics
comparable to human memories [6]. For simplicity, here, we further
abstract this complex network model by the methods of mean-field
theory. Thereby, the dynamics of each cell assembly in the network are
reduced to two differential equations; one for the average neuronal
activity and one for the average synaptic weight. Given this simplified
model, we show that, in contrast to loosely connected groups of neurons,
the formation of cell assemblies lets the activity of the neuronal
population follow a hysteresis which creates complex network dynamics,
which depend on the initial conditions. Given this hysteresis, in the next
step, we connect two such assemblies with each other by plastic
connections also adapted by the interaction between synaptic plasticity
and scaling. Amongst others, we analyzed under which circumstances the
dynamics of this two-cell-assembly system are comparable to the above
mentioned dynamics of human and animal memories and how different
parameters of the system (e.g., cell assembly size) influence them.
In summary, this work is one of the first using a mathematical model to
relate the dynamics of memories on a psychological scale to the
hypothesized dynamics of cell assemblies on a neuronal scale.
References
1. Eichenbaum H: The cognitive neuroscience of memory: An introduction.
Oxford University Press 2012.
2. Martin SJ, Grimwood PD, Morris RGM: Synaptic plasticity and memory: An
evaluation of the hypothesis. Annual Review Neuroscience 2000, 23:649-711.
3. Hebb DO: The Organization of Behaviour. Wiley, New York 1949.
4. Tetzlaff C, Kolodziejski C, Timme M, Wrgtter F: Synaptic scaling in
combination with many generic plasticity mechanisms stabilizes circuit
connectivity. Frontiers in Computational Neuroscience 2011, 5:47.
5. Turrigiano GG, Leslie KR, Desai NS, Rutherford LC, Nelson SB: Activity-
dependent scaling of quantal amplitude in neocortical neurons. Nature
1998, 391:892-896.
6. Tetzlaff C, Kolodziejski C, Timme M, Tsodyks M, Wrgtter F: Synaptic
scaling enables dynamically distinct short- and long-term memory
formation. PLoS Computational Biology 2013, 9(10):e1003307.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
P254
Towards a biological plausible model of the interaction of long-term
memory and working memory
Timo Nachstedt1,2*, Florentin Wrgtter1,2, Christian Tetzlaff1,2
1
Third Institute of Physics, Georg-August-Universitt, Gttingen, 37077,
Germany; 2Bernstein Center for Computational Neuroscience, Gttingen,
37077, Germany
E-mail: timo.nachstedt@phys.uni-goettingen.de
BMC Neuroscience 2015, 16(Suppl 1):P254
Learning and memorizing past information are critical features of neural
networks. Several models have been proposed to explain the storage
and retrieval of different kinds of information [1,2] . These models differ
in terms of the types of memories they can store, in terms of the time
scales on which they operate, and in terms of the persistence of the
stored information. In particular, there are neural models for the long-
term memory (LTM) system [2] and other neural models for the
working memory (WM) system [1]. While there are some controversies
whether LTM and WM are realized by distinct brain regions, there is no
doubt about the fact that these two systems are at the same time
tightly coupled and interact continuously [3]. However, it is still widely
unclear how this interaction is actually realized. Here, we demonstrate a
model reproducing and explaining the interaction between LTM and
WM such as to exploit the abilities and advantages of the individual
components.
WM is most probably realized in the prefrontal cortex (PFC). To capture
the diverse dynamics observed in PFC during WM experiments, we
choose to model WM by a reservoir computing network [4]. Basically, it
consists of a large reservoir of randomly and recurrently connected
neurons coupled to the input signals. Due to the huge variety of different
signals present in the reservoir, these networks can serve as universal
function approximators. However, the influence of past inputs fades away
within time scales of seconds to minutes.
The hippocampus, especially its CA3 area, is believed to play a major
role in LTM formation on time scales of hours to days. We model its
auto-associative capabilities by the well-known idea of cell assemblies
[5]. According to this idea, declarative knowledge translates into
correlations of neural activities which, in turn, lead to a potentiation of
excitatory synapses. This process forms groups of highly interconnected
neurons - named cell assemblies. If the synapses in a recurrent network
are governed by Hebbian plasticity, cell assemblies emerge in an
unsupervised manner based on correlations observed in the network
input. Experimental findings indicate that rather abstract or general
LTM knowledge does not depend on the hippocampus but is
transferred to the medial prefrontal cortex (mPFC). In our model, the
representations in the mPFC are the central interface between LTM and
WM. It receives signals from both the cell-assembly as well as the
reservoir network.
Analyzing the capabilities and characteristics of the proposed system, we
train the reservoir to perform certain arithmetic operations within
different contexts. During operation, the cell assembly network learns to
detect and remember associations between context signals and
computed results. This enables the system to remember results of earlier
calculations beyond the WM capacity and to reuse them for later tasks.
Thus, the system is able to perform computations based on short-term
memories of the operands as well as to overcome the limits of the short-
term memory by storing computed results into the long-term memory
and transferring them back for further usage.
References
1. Baddeley A: Working Memory: Theories, Models, and Controversies. Annu.
Rev. Psychol 2012, 63:1-29.
2. Wood R, et al: A review of long-term memory in natural and synthetic
systems. Adaptive Behavior 2011, 20(2):81-103.
3. Jonides J, et al: The Mind and Brain of Short-Term Memory. Annu. Rev.
Psychol 2008, 59:193-224.
4. Lukosevicius M, Jaeger H: Reservoir Computing Approaches to Recurrent
Neural Network Training. Comput. Sc. Rev 2009, 3(3):127-149.
5. Hebb D: The Organization of Behavior New York: Wiley 1949.
Page 166 of 200
P255
Sparse coding and dictionary learning for spike trains to find spatio-
temporal patterns
Taro Tezuka
Faculty of Library, Information and Media Science, University of Tsukuba,
Tsukuba, 305-0821, Japan
E-mail: tezuka@slis.tsukuba.ac.jp
BMC Neuroscience 2015, 16(Suppl 1):P255
In biological neural networks, it is widely accepted that the spikes are the
fundamental building blocks of information representation [1]. In contrast,
whether such building blocks exist at a higher level in terms of time and in a
population of neurons is a topic of ongoing debate. One approach for finding
candidates for such building blocks is to seek for frequently appearing spike
patterns in a population. These sequences are often called spatio-temporal
patterns, cell assemblies, or unitary events [2-4]. They could metaphorically be
considered as an alphabet of neural information processing [5,6]. Some
patterns have already been found and are related to functional roles such as
memory consolidation and gating of sensory inputs [7,8].
One difficulty in finding spatio-temporal patterns arises from observed
spike trains being a superposition of multiple patterns. In signal
processing, one commonly used method for decomposing the signal into
patterns is dictionary learning for sparse coding [9-11]. Sparse coding
expresses the input signal as a linear combination of a few template
vectors taken from a matrix called a dictionary or codebook. In terms of
linear algebra, sparse coding corresponds to finding a sparse vector x,
which fulfills y = Dx, where y is the observed signal vector and D is a
dictionary. When the dimension of is much larger than that of y, it is
possible to find sparse x. Each column of D is called an atom, which
represents a template vector. A good dictionary decomposes the most of
the observed signals into a small set of template vectors. In other words, D
must sparsify not just one input vector y but many others as well. This is
represented by using matrix Y whose column vectors are observed signals.
In this case, sparse coding is represented by equation Y = DX. The goal is
to find sparse matrix given Y and D. Whether input matrix Y can be
transformed into sparse or not depends on dictionary D. The goodness
of D depends on Y. The task of finding optimal D given Y is called
dictionary learning. In this work sparse coding and dictionary learning
were applied for finding spatio-temporal patterns from multivariate spike
trains. Spike trains were transformed to vectors using binning, that is,
converted to vectors of short-time firing rates. The methods were tested
using different bin sizes. The results obtained for biological data showed
possible candidates of spatio-temporal patterns in neural activity.
Acknowledgement: This work was supported in part by JSPS KAKENHI
Grant Numbers 21700121, 25280110, and 25540159.
References
1. Gerstner W, Kistler WM, Naud R, Paninski L: Neuronal Dynamics Cambridge:
Cambridge University Press 2014.
2. Villa AEP: Empirical evidence about temporal structure in multi-unit
recordings. Conceptual Advances in Brain Research 2000, 3:1-51.
3. Buzsaki G: Neural syntax: cell assemblies, synapsembles, and readers.
Neuron 2010, 68(3):362-385.
4. Grun S, Diesmann M, Aertsen A: Unitary events in multiple single-neuron
spiking activity. I. Detection and significance. Neural Computation 2002,
14:43-80.
5. Baram Y: Global attractor alphabet of neural firing modes. Journal of
Neurophysiology 2013, 110:907-915.
6. Eyherabide HG, Samengo L: The information transmitted by spike
patterns in single neurons. Journal of Physiology Paris 2010, 104:147-155.
7. Nadasdy Z, Hirase H, Czurko A, Csicsvari J, Buzsaki G: Replay and time
compression of recurring spike sequences in the hippocampus. Journal
of Neuroscience 1999, 19(21):9497-9507.
8. Luczak A, Bartho P, Harris KD: Gating of sensory input by spontaneous
cortical activity. Journal of Neuroscience 2013, 33(4):1684-1695.
9. Hoyer PO: Non-negative matrix factorization with sparseness constraints.
Journal of Machine Learning Research 2004, 5:1457-1469.
10. Bruckstein A, Donoho DL, Elad M: From sparse solutions of systems of equations
to sparse modeling of signals and images. SIAM Review 2009, 51(1):34-81.
11. Elad M: Sparse and Redundant Representations Berlin: Springer 2010.
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 167 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

different implementations of RC display very little resistance to small

P256
Extreme sensitivity of reservoir computing to small network disruptions
Philippe Vincent-Lamarre1*, Guillaume Lajoie2,3, Jean-Philippe Thivierge1
1
School of Psychology and Center for Neural Dynamics, University of Ottawa,
Ottawa, Ontario K1N 6N5, Canada; 2UW Institute for Neuroengineering,
University of Washington, Seattle, WA, US; 3Max Planck Institute (DS) and
Bernstein Center for Computational Neuroscience, Gttingen, Germany
E-mail: pvinc058@uottawa.ca
BMC Neuroscience 2015, 16(Suppl 1):P256
Recent computational models based on reservoir computing (RC) are
gaining attention as plausible theories of cortical information
processing. In these models, the activity of a recurrently connected
population of neurons is sent to one or many read-out units through a
linear transformation. These models can operate in a chaotic regime
which has been proposed as a possible mechanism underlying
sustained irregular activity observed in cortical areas [1,2]. Furthermore,
models based on RC replicate the neural dynamics involved in decision
making [3], interval timing [2], and motor control [1]. However, one
biological constraint that has been overlooked in these models is their
resistance to small connectivity perturbations such as failures in synaptic
transmission, a phenomenon that occurs frequently in healthy circuits
without causing any drastic functional changes. Here, we show that
synaptic disruptions and discuss the implications of such fragility for RC
mechanisms that may be present in neural coding. With the FORCE [1]
procedure, networks lost their ability to replicate a jagged sinusoidal
signal after a single neuron was removed from the reservoir (Figure 1A).
Networks with innate training [2] showed a similar effect on a timing
task (Figure 1B). The lag in the timing and the noise in the output both
increased monotonically as further neurons were removed (Figure 1C,D);
networks reached random performance after ~1.5% of neurons were
eliminated. After the suppression of a single neuron, the spectrum of
the weight matrix was greatly disturbed and repeated trials displayed
unreliable trajectories, as assessed with principal components analysis.
When individual synapses were removed instead of neurons, networks
reached random performance after ~0.5% of synapses from the
reservoir were eliminated. While living neuronal circuits can withstand
small synaptic disruptions without compromising task performance, our
results suggest that such disruptions have a catastrophic impact on the
behaviour of RC models. Retraining the read-out unit seems to be futile
as it results as a completely new solution post retraining instead of a
finer restructuration. These results cast doubt on the validity of a large
class of models that claim to capture the neuronal mechanisms of
cognitive and behavioral tasks.
Acknowledgements: This research was funded by grants to J.P.T. from
NSERC Discovery and CIHR operating funds.

Figure 1(abstract P256) Performance of damaged reservoirs of 1,000 neurons with FORCE and innate learning algorithms. A. Target signal
(green, perfectly replicated with the originally trained network) and the trace of the same network after the removal of one neuron in its reservoir. B. Ten
trials (red) with different initial conditions of a damaged network (N-2 neurons) that is trained to peak at 1,000 ms (green) using innate learning. C.
Average lag between the target and the output timing (100 trials per condition) as a function of the number of removed neurons. D. Mean squared error
as a function of the number of removed neurons
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
References
1. Sussillo D, Abbott LF: Generating Coherent Patterns of Activity from
Chaotic Neural Networks. Neuron 2009, 63:544-557.
2. Laje R, Buonomano DV: Robust timing and motor patterns by taming
chaos in recurrent neural networks. Nat Neurosci 2013, 16:925-933.
3. Barak O, Sussillo D, Romo R, Tsodyks M, Abbott LF: From fixed points to
chaos: Three models of delayed discrimination. Progress in Neurobiology
2013, 103:214-222.
P257
Role of Na+ and Ca2+ currents in computational model of in-vitro sigh
generation
Natalia Toporikova1*, Muriel Thoby-Brisson2
1
Biology Department and Neuroscience Program, Washington and Lee
University, Lexington, VA, 24450, USA; 2Institut de Neurosciences Cognitives
et Intgratives dAquitaine, CNRS UMR 5287, Universit de Bordeaux, 33076
Bordeaux, France
E-mail: toporikovan@wlu.edu
BMC Neuroscience 2015, 16(Suppl 1):P257
Eupneic breathing in mammals is periodically interrupted by spontaneous
augmented breaths (sighs) that are characterized by a biphasic
larger-amplitude inspiratory burst followed by post-sigh apnea. Previous in
vitro studies in newborn rodents have demonstrated that the respiratory
oscillator of the pre-Btzinger complex (preBtC) can generate the distinct
inspiratory-related motor patterns for both eupnea- and sigh-like activity
[1,2]. However it remains debated whether these two types of inspiratory
activities are produced by the same neuronal population or by distinct sub-
networks. Based on recent in vitro data obtained in the mouse embryo [3],
we have built a computational model consisting of two compartments, one
dedicated to sigh generation and the other generating eupneic bursts,
interconnected through appropriate synapses (Figure 1 A).
The model reproduces basic features of simultaneous sigh and eupnea
generation: two types of bursts differing in terms of shape, amplitude and
frequency of occurrence and mimics the effect of glycinergic synapses
blockade. We designed a two-compartment computational model for sigh
and eupnea subpopulations of neurons with several different parameters
reflecting distinct burst generating mechanisms. The sigh subpopulation
generates a low frequency rhythm based on slow intracellular Ca 2+
oscillations and the eupnea subnetwork generates fast oscillations mainly
driven by activation/inactivation of the persistent Na+ current (Fig 1 B,C).
Furthermore, we used this model to make predictions that were
subsequently tested on the isolated preBtC in brainstem slice preparations.
Through a combination of our in vitro and in silico approaches we found
that 1), sigh events are less sensitive to network excitability than eupneic
activity, 2)The combination of voltage-gated calcium current and persistent
sodium current control the sigh period of, and 3), specific parameters of Ih
activation set the low sensitivity to excitability in the sigh neuronal subset.
Altogether, our results strongly support the hypothesis that distinct
subpopulations within the preBtC network are responsible for sigh and
eupnea rhythmogenesis.
References
1. Lieske SP, Ramirez J-M: Pattern-specific synaptic mechanisms in a
multifunctional network. I. Effects of alterations in synapse strength.
Journal of Neurophysiology 2006, 95(3):1323-1333.
2. Tryba AK, Pea F, Lieske SP, Viemari J-C, Thoby-Brisson M, Ramirez J-M:
Differential modulation of neural network and pacemaker activity
underlying eupnea and sigh-breathing activities. Journal of
Neurophysiology 2008, 99(5):2114-2125.
3. Chapuis C, Autran S, Fortin G, Simmers J, Thoby-Brisson M: Emergence of
sigh rhythmogenesis in the embryonic mouse. The Journal of Physiology
2014, 592(Pt 10):2169-2181.
P258
Fast and accurate representations of stochastic ion channel fluctuations
David F Anderson1, Bard Ermentrout2, David D Friel3, Roberto F Galn3,
Benjamin Lindner4,5, Shusen Pu6, Deena R Schmidt7, Peter J Thomas6*
1
Department of Mathematics, University of Wisconsin, Madison, WI 53706,
USA; 2Department of Mathematics, University of Pittsburgh, Pittsburgh, PA
15260, USA; 3Department of Neurosciences, Case Western Reserve University,
Page 168 of 200
Cleveland, OH 44106, USA; 4Bernstein Center for Computational
Neuroscience, 10115 Berlin, Germany; 5Department of Physics, Humboldt
University, 12489 Berlin, Germany; 6Department of Mathematics, Applied
Mathematics and Statistics, Case Western Reserve University, Cleveland, OH
44106, USA; 7Department of Mathematics and Statistics, University of
Nevada, Reno, NV 89557, USA
E-mail: pjthomas@case.edu
BMC Neuroscience 2015, 16(Suppl 1):P258
Random ion channel gating is an important source of noise at the single
neuron level. For mesoscale ion channel population sizes, fast, accurate
representation of channel noise fluctuations remains an important
challenge. We highlight recent progress in three areas.
In [1] we present an exact stochastic simulation algorithm, which takes into
account the time dependence of the voltage sensitive transitions due to
rapid voltage changes during action potentials. The exact algorithm is similar
to a widely used approximate stochastic simulation algorithm, in which
transition rates are held fixed during the intervals between channel state
transitions. We compare the algorithms and show that they are inequivalent
in a strong sense, meaning that sample paths diverge when driven with
identical Poisson processes, leading to different precise firing times. But the
stationary histograms produced are practically indistinguishable for modest
channel numbers (circa N>100), indicating weak equivalence.
For channel numbers in the hundreds or higher, numerical stochastic
differential equations (SDE) algorithms based on the system size expansion
can be significantly faster than simulations based on discrete Markov chain
simulations, while retaining reasonable accuracy. For large networks,
however, even SDE based simulations become costly, particularly as more
complex channel gating schemes are introduced. Schmandt and Galn
introduced a stochastic shielding approximation as a fast, accurate way of
simulating stochastic ion channel kinetics [2]. In the SDE representation,
each edge in the graph generates both a mean population flux between
adjacent nodes, and a fluctuation about the mean. Only the fluctuations
arising from edges connecting functionally distinct states directly affect
fluctuations in the observed behavior of the cell. In [3] we analyze
stochastic shielding both for the HH sodium and potassium gating models,
and for an ensemble of random graphs. We derive a quantitative measure
of edge importance related to the eigenvalue/eigenvector decomposition
of the graph Laplacian matrix.
Channel noise makes a regularly spiking neuron a stochastic oscillator.
The classical definition for the asymptotic phase of an oscillator breaks
down when stochasticity is taken into account. Alternative definitions of
the phase have been based on the mean first passage time property of
a system of isochronal surfaces [4] and in terms of the spectral
decomposition of a the adjoint Kolmogorov operator [5,6]. In the
vanishing noise limit, the spectrally derived isochrons approach those of
the underlying mean field system, if the latter has a finite period limit
cycle.
Together, these results expand our analytic, numerical, and conceptual
tools for understanding the effects of random ion channel gating in
conductance based neural models.
Acknowledgements: Simons Foundation (#259837 to PJT), Council for
the International Exchange of Scholars (CIES), Mathematical Biosciences
Institute (NSF-DMS 0931642), NSF EF-1038677, DMS-1413770, DMS-
1318832, DMS-1219754, Bundesministerium fr Bildung und Forschung
(FKZ: 01GQ1001A), Army Research Office W911NF-14-1-0401.
References
1. Anderson David F, Bard Ermentrout, Peter J Thomas: Stochastic
representations of ion channel kinetics and exact stochastic simulation
of neuronal dynamics. J Comput Neurosci 2014, 38(1):67-82.
2. Schmandt Nicolaus T, Roberto F Galn: Stochastic-shielding approximation
of Markov chains and its application to efficiently simulate random ion-
channel gating. Phys rev lett 2012, 109(11):118101.
3. Schmidt Deena R, Peter J Thomas: Measuring edge importance: a
quantitative analysis of the stochastic shielding approximation for
random processes on graphs. J Math Neurosci 2014, 4(1):6.
4. Schwabedal Justus TC, Arkady Pikovsky: Phase description of stochastic
oscillations. Phys rev lett 2013, 110(20):204102.
5. Thomas Peter J, Benjamin Lindner: Asymptotic Phase for Stochastic
Oscillators. Phys rev lett 2014, 113(25):254101.
6. Potoyan Davit A, Peter G Wolynes: On the dephasing of genetic
oscillators. Proc Natl Acad Sci 2011, 111(6):2391-2396.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P257) Sigh and eupnea activity pattern in silico. (A) Diagrams of sigh (left) and eupnea (right) network models. Except for the ER
capacity, INaP and Ih, the sigh and eupnea models are identical. (B) Intracellular Ca2+ (top) and voltage (bottom) obtained for individual uncoupled
compartments (gsyn=0)
P259
Mechanisms for synchronized burst firing in pyramidal cells using
oscillatory inhibition: a model for attentional control
Caroline Fischer*, Paul HE Tiesinga, Marije ter Wal
Department of Neuroinformatics, Donders Centre for Neuroscience, Radboud
University Nijmegen, 6525 AJ, Nijmegen, the Netherlands
E-mail: caroline.fischer@student.ru.nl
BMC Neuroscience 2015, 16(Suppl 1):P259
Pyramidal cells have two different action potential initiation sites; one at
the soma, which receives inputs from synapses contacting the basal
dendrites or the soma, and one at the distal apical dendrite and tuft [1].
At the latter, calcium spikes can be elicited resulting in a plateau
Page 169 of 200
potential due to the influx of calcium that causes the pyramidal cell to
fire bursts of action potentials [2]. This mechanism can be used to
enhance the effect of top-down synaptic inputs that are received on the
distal apical dendrite and tuft on activating its downstream targets.
Calcium spikes can be elicited either by convergence of forward-
propagating inputs at the tuft or by simultaneous depolarization from
back-propagating action potentials and excitation at the tuft [3], referred
to as FAC and BAC, respectively.
We investigated whether these two firing patterns can be synchronized by
oscillatory inhibition in order to transmit top-down or combined top-down
and bottom-up information to different brain areas. We hypothesize this
mechanism as an explanation for the synchronization of burst spiking to
distant local field potentials at beta frequencies (12-20 Hz) that was
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
recorded in the prefrontal cortex and the anterior cingulate cortex during
attentional control [4].
We simulated the behavior of pyramidal cells stimulated by oscillatory
inhibition at the tuft and noisy excitatory inputs at the basal dendrite and
tuft. First, we investigated the occurrence and phase-locking of FAC and
BAC firing in one cell as function of the ratio of excitatory basal and tuft
inputs. Second, we examined the synchronization of a population of
pyramidal cells downstream to oscillatory inhibition with frequencies in
the beta range. The pyramidal cells were modeled using four
compartments and including calcium dynamics in the apical dendrite and
tuft [5].
For single pyramidal cells FACs predominantly occur when the tuft
receives stronger excitation than the basal dendrites, while BACs occur
predominantly when the tuft and basal dendrites are excited equally
strong. Both mechanisms have in common that they dominate the
behavior of the pyramidal cell only when oscillatory inhibition is in the
lower frequency range (2-20 Hz). For a 1:2 ratio between excitatory basal
and tuft input FACs accounted for up to 85% of all spike events, whereas
for a ratio of 1:1 BACs constituted up to 68%. Bursts activated by FACs
and BACs were strongly phase-locked to the inhibitory drive, while bursts
activated by other mechanisms had a lower coherence. The synchrony
between pairs of pyramidal cells was quantified using the Schreiber
measure [6] with the smoothing parameter set to s=5 ms. When
stimulated in the beta frequency range the synchrony between pairs of
pyramidal cells was on average 0.78. Taken together, the results show
that FAC and BAC firing can be used in conjunction with oscillatory
inhibition to produce synchronized burst firing in pyramidal cells. This
provides an effective means to transmit top-down or combined top-down
and bottom-up information downstream.
Acknowledgements: This work is supported by funding from the
European Unions Seventh Framework Programme (FP7/2007-2013) under
grant agreement no. 600925.
References
1. Yuste R, Gutnick MJ, Saar D, Delaney KR, Tank DW: Ca2+ Accumulations in
Dendrites of Neocortical Py-ramidal Neurons: An Apical Band and
Evidence for Two Functional Compartments. Neuron 1994, 13:23-43.
2. Schiller J, Schiller Y, Stuart G, Sakmann B: Calcium action potentials
restricted to distal apical dendrites of rat neocortical pyramidal neurons.
J Physiol 1997, 505(3):605-616.
3. Larkum ME, Zhu JJ, Sakmann B: A new cellular mechanism for coupling
inputs arriving at different cortical layers. Nature 1999, 398:338-341.
4. Womelsdorf T, Ardid S, Everling S, Valiante TA: Burst Firing Synchronizes
Prefrontal and Anterior Cingulate Cortex during Attentional Control.
Current Biology 2014, 24:2613-2621.
5. Bahl A, Stemmler MB, Herz VM, Roth A: Automated optimization of a
reduced layer 5 pyramidal cell model based on experimental data.
J Neurosci Methods 2012, 210:22-34.
6. Kreuza T, Haas JS, Morellic A, Abarbanel HDI, Politi A: Measuring spike train
synchrony. J Neurosci Methods 2007, 165:151-16.
P260
Alpha phase modulates the effectiveness and directionality of cortical
communication
Silvan C Quax*, Paul Tiesinga
Neuroinformatics department, Radboud University, Nijmegen, the
Netherlands
E-mail: s.c.quax@student.ru.nl
BMC Neuroscience 2015, 16(Suppl 1):P260
The selective routing of information between cortical areas is important for
efficient communication in the brain. Recent experiments have shown an
increase of alpha band coherence with selective attention between the
pulvinar, V4, and TEO [1]. Strong cross-frequency coupling between
oscillations in the alpha and gamma band [2] supports the idea that the
alpha rhythm coordinates communication between higher frequency
oscillations. More evidence has shown that alpha phase adjustments occur
during an attentional distractor task [3]. It remains unknown however, how
the alpha phase could influence selective communication. Here we
investigated whether shifting the relative alpha phase between two
cortical areas could coordinate cortical communication.
Page 170 of 200
A network model was constructed comprised of two neuronal
populations, each representing a cortical area, with each neuron modeled
using the Izhikevich model [4]. Parameters were chosen such that stable
gamma oscillations emerged, whereas the alpha rhythm was
implemented as a periodic modulation of the input to both areas. We
modulated the amplitude and relative phase of the alpha rhythm in order
to investigate their effects on cortical communication, which was
quantified as the coherence in the gamma frequency band.
Results showed that the relative phase of the alpha modulation between
the two neuronal populations strongly affected their gamma coherence
(Figure 1A). The modulation depth of the gamma coherence increased
with higher alpha amplitude (Figure 1B). The relative alpha phase also had
effect in a model with recurrent connectivity between the neuronal
populations, where it determined the directionality of communication
between the populations. Finally, the alpha phase of a neuronal
population modulated its response to an external synaptic input
representing a stimulus. When the relative alpha phase between the
neuronal populations was optimal, a stimulus had a bigger impact on the
second cortical area.
These results indicate that the relative alpha phase between neuronal
populations strongly influences the effectiveness and directionality of
their communication. This suggests that, during selective attention, the
brain could be actively manipulating the relative phase of the alpha
modulation between different cortical areas in order to coordinate the
effectiveness of communication and the balance between feedforward
and feedback communication between cortical areas. A prime candidate
for coordinating phase shifts in the alpha rhythm is the pulvinar [1].
Taken together, our results show that the brain could coordinate cortical
communication by dynamically changing the relative alpha phase
between cortical areas. However, experimental manipulation of the
relative alpha phase is necessary to validate this conclusion and to clarify
the role of the pulvinar in this mechanism.
Acknowledgements: The research leading to these results has received
funding from the European Union Seventh Framework Programme (FP7/
2007-2013) under grant agreement 600925 (NEUROSEEKER).
References
1. Saalmann YB, Pinsk MA, Wang L, Li X, Kastner S: The pulvinar regulates
information transmission between cortical areas based on attention
demands. Science 2012, 337(6095):753-756.
2. Osipova D, Hermes D, Jensen O: Gamma power is phase-locked to
posterior alpha activity. PLoS One 2008, 3(12):e3990.
3. Bonnefond M, Jensen O: Alpha oscillations serve to protect working
memory maintenance against anticipated distracters. Curr biol 2012,
22(20):1969-1974.
4. Izhikevich EM: Simple model of spiking neurons. IEEE Trans Neural Netw
2003, 14(6):1569-1572.
P261
Analysis of replacing DNase-seq data with histone marks in
computational dimer prediction
Victor Chukwudi Osamor1,2*, Jerzy Tiuryn2
1
Department of Computer and Information Sciences, Covenant University, P.
M.B 1023, Ota, Ogun State, Nigeria; 2Institute of Informatics, University of
Warsaw, ul Banacha 2, 02-097, Warsaw, Poland
E-mail: vcosamor@gmail.com
BMC Neuroscience 2015, 16(Suppl 1):P261
Background: Open chromatin regions and their findings are great pointers
to genome-wide revelations of transcription factor activities, which in-turn
defines the level of gene participation and their roles in metabolic well-
being or disease condition in human body. Chromatin openness [1] is highly
influenced by the presence of histone markers which are biomarkers that
define the selective modification of some amino acid at specific positions.
Worthy of note is that DNase-seq useful in detecting transcription [2] activity
is scarcely available in most species compared to more abundant ChIP-seq
data across several species. Our interest is to investigate an optimal
combination of histone marks that could replace the DNase-seq data in
transcription factor dimer prediction algorithm.
Methods: The experimental design of this work involves the analysis of
the DNase-seq and all possible combinations of 5 corresponding histone
markers (H3k4me1, H3k4me2, H3k4me3, H3k9ac and H3k27ac) across 3
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P260) Alpha phase influences gamma coherence. (A) Coherence in the gamma band between the two neuronal populations is
strongly modulated by the relative alpha phase between the two populations. (B) The difference in gamma coherence between best alpha phase (-45,
red line) and worst alpha phase (135, blue line), increases with alpha amplitude
human cell lines namely GM12878, H1hESC and K562 for a total of
31 experimental set-up. The BAM files are applied to Model-based
Analysis of ChIP-Seq (MACS) [3] via its variant MACS14 to call the MACS
peaks of the histone modification which is combined, sorted, merged and
grouped according to base pair length not greater than thresholds
starting from initial 500bp, 1000bp to 30000bp by increment of 1000bp.
We are considering two kinds of regime of working with Transcription
Factor (TF) dimer prediction algorithms called Transcription factor
Association for Complex Overrepresentation (TACO) [4]: (1) Strongly cell-
type specific (SCTS) which refers to genomic areas with chromatin
opened specifically only for the considered cell-type and closed in all
other cells. (2) Weakly cell-type specific (WCTS) where we consider all
open area in a given cell-type without reference to other cell-types. These
prepared MACS peaks data are clustered in TACO supported by
29 experimentally proven TF dimers. The narrowPeaks of DNase-seq were
directly applied to TACO without any need to call their peaks or set
threshold but as gold standard to assess the quality of all possible
combinatorial effects [5] of histone marks under investigation.
Results: On assessing how much quality prediction we may be
compromising when we substitute DNase-seq data with histone
modification in TACO, we discovered that a total of 7 experimentally
proven dimers in literature were predicted correctly by TACOs WCTS
algorithm in the ratio of 4:2:1 for GM12878, K562 and H1hESC cell-types
respectively. This is lower than 9 experimentally proven dimers in literature
obtained by our best WCTS prediction in Experiment 10 for a combination
of histone modifications involving H3k4me1 and H3k9ac over 5000-
7000bp. In addition, using TACOs SCTS algorithm with DNase-seq
recorded 8 experimentally proven dimers in the ratio of 5:2:1 for GM12878,
K562 and H1hESC respectively, while a total of 12 experimentally proven
dimers were predicted in the ratio of 2:7:3 for the corresponding cell-types
trimethylation (H3k4me3) in Experiment 3.
Conclusions: Interestingly, our result indicates that the suitable basepair
range for an optimal substitution of DNase-seq with histone marks for
dimer prediction may possibly be in the range of 5000-7000bp with
5000bp being the preferred. Furthermore, single modification seem to
favour SCTS prediction while the WCTS tend to be more relevant for
combined modifications. Our method for replacing DNase-seq with
histone appears to have numerically out-performed DNase-seq and could
Page 171 of 200
further help in discovering chromatin openness in brain cells as well as
other human cell line as applied in embryonic stem cell that could
differentiate into neuro-sensory cells and organs.
Acknowledgements: VCO conducted this research while visiting
University of Warsaw as a Postdoc Marie Curie/ERCIM Fellow. The
research leading to these results has received funding from the European
Union Seventh Framework Programme (FP7/2007-2013) under grant
agreement n 246016.
References
1. Cui P, Li J, Sun B, Zhang M, Lian B, Li Y, Xie L: A Quantitative Analysis of
the Impact on Chromatin Accessibility by Histone Modifications and
Binding of Transcription Factors in DNase I Hypersensitive Sites. BioMed
Res Int 2013, 2013:e914971.
2. Sherwood RI, Hashimoto T, ODonnell CW, Lewis S, Barkal AA, van Hoff JP,
Karun V, Jaakkola T, Gifford DK: Discovery of directional and
nondirectional pioneer transcription factors by modeling DNase profile
magnitude and shape. Nat Biotechnol 2014, 32:171-178.
3. Zhang Y, Liu T, Meyer CA, Eeckhoute J, Johnson DS, Bernstein BE,
Nusbaum C, Myers RM, Brown M, Li W, Liu XS: Model-based Analysis of
ChIP-Seq (MACS). Genome Biol 2008, 9:R137.
4. Jankowski A, Prabhakar S, Tiuryn J: TACO: a general-purpose tool for
predicting cell-type-specific transcription factor dimers. BMC Genomics
2014, 15:208.
5. Karch KR, DeNizio JE, Black BE, Garcia BA: Identification and interrogation
of combinatorial histone modifications. Epigenomics Epigenetics 2013,
4:264.
P262
Fitness and neural complexity of animats exposed to environmental
change
Larissa Albantakis*, Giulio Tononi
Department of Psychiatry, University of Wisconsin, Madison, WI, 53719, USA
E-mail: albantakis@wisc.edu
BMC Neuroscience 2015, 16(Suppl 1):P262
We recently showed that adaptive logic-gate networks (animats) increase
their capacity to integrate information as they adapt to environments of
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
increasing complexity [1]. The animats task environments consisted of
falling blocks of different sizes, which had to be either caught or avoided.
When the animats were exposed to more difficult task environments that
required more internal memory, they developed more complex networks
(brains), as indicated by a larger number of irreducible internal
mechanisms (concepts) and higher integrated conceptual information
(F) [2,3]. Animats with brains of high F outperformed animats with
modular or feedforward brains because they could pack a larger number
of mechanisms for the same number of nodes and connections. Here we
investigate whether this key feature of animats of high F leads to greater
flexibility in adapting to environmental changes. We selected animats
with integrated (F > 0) or modular (F = 0) structures that had adapted
perfectly, within 30,000 generations, to a particular task environment. We
then enhanced the difficulty of the task environment by adding blocks of
various sizes that were to be caught or avoided in a way that was either:
i) congruent with the old task, ii) neutral to the old task, or iii)
incongruent with the old task, reversing the previous rules. For congruent
changes in the task environment, the fitness of animats with integrated
brains and many irreducible mechanisms, as compared to modular
animats, (1) dropped less right after the change, (2) recovered faster, and
(3) maintained higher values even after 30,000 generations in the new
environment. These results corroborate the hypothesis that brains with
high capacity for information integration may provide an evolutionary
advantage in complex, changing environments.
Acknowledgements: This work has been supported by the Templeton
World Charities Foundation (Grant #TWCF 0067/AB41).
References
1. Albantakis L, Hintze A, Koch C, Adami C, Tononi G: Evolution of Integrated
Causal Structures in Animats Exposed to Environments of Increasing
Complexity. PLoS Comp Biol 2014, 10(12):e1003966.
2. Oizumi* M, Albantakis* L, Tononi G: From the Phenomenology to the
Mechanisms of Consciousness: Integrated Information Theory 3.0. PLoS
Comp Biol 2014, 10(5):e1003588.
3. Tononi G: Integrated information theory. Scholarpedia 2014, 10(1):4164.
P263
Stochastic mean-field formulation of the dynamics of diluted neural
networks
David Angulo-Garcia*, Alessandro Torcini
Istituto dei Sistemi Complessi, Consiglio Nazionale delle Ricerche (CNR), via
Madonna del Piano 10, Sesto Fiorentino, Italy I-50019
E-mail: david.angulo@fi.isc.cnr.it
BMC Neuroscience 2015, 16(Suppl 1):P263
We consider pulse-coupled leaky integrate-and-fire neural networks with
randomly distributed synaptic couplings. This random dilution induces
fluctuations in the evolution of the macroscopic variables and
deterministic chaos at the microscopic level [1]. Our main aim is to mimic
the effect of the dilution as a noise source acting on the dynamics of a
Figure 1(abstract P263) A. Inter-spike interval distribution functions and B. attractor dynamics for a representative case of dilution with system
size ND = 5000. C. Maximal Lyapunov exponent as a function of network size. In all panels, black circled symbols indicate the diluted deterministic
network, blue triangles and red squares denote the stochastic mean-field approximation obtained by employing white and colored noise respectively.
For all the panels the deterministic systems are diluted down to the 80%, while the fully coupled stochastic systems are always composed of only
100 neurons
Page 172 of 200
globally coupled nonchaotic system. We show that the evolution of a
deterministic diluted neural network of any size can be well approximated
by a much smaller fully coupled network, where each neuron is driven by
a mean synaptic current plus additive noise. These terms represent
the average and the fluctuations of the synaptic currents acting on the
single neurons in the diluted system. The main microscopic and
macroscopic dynamical features can be reproduced within this stochastic
approximation. In order to illustrate the quality of this reconstruction, we
compare the probability distribution function of the inter-spike intervals
and the macroscopic attractor of the deterministic diluted system with
those obtained by employing the mean-field stochastic model (see Figure
1A and 1B, respectively). Furthermore, the microscopic stability of the
diluted network can be also reproduced, as demonstrated from the almost
coincidence of the measured Lyapunov exponents in the deterministic and
stochastic cases for an ample range of system sizes (see Figure 1C.). Our
results strongly suggest that the fluctuations in the synaptic currents are
responsible for the emergence of chaos in this class of pulse-coupled
systems [2].
Acknowledgements: This work has been supported by the European
Commission under the program Marie Curie Network for Initial Training,
through Project No. 289146, Neural Engineering Transformative
Technologies (NETT). D.A.-G. also acknowledges the partial support provided
by Departamento Adminsitrativo de Ciencia Tecnologia e Innovacion-
Colciencias through the program Doctorados en el exterior-2013.
References
1. Olmi S, Livi R, Politi A, Torcini A: Collective oscillations in disordered
neural network. Phys. Rev. E 2010, 81(4):046119.
2. Angulo-Garcia D, Torcini A: Stochastic mean-field formulation of the
dynamics of diluted neural networks. Pyhs. Rev. E in press, doi:10.1103/
PhysRevE.00.002900.
P264
Two different mechanisms alternate during cortical synchronized states
Erin Munro1*, Tansi Khodai2, Shuzo Sakata2, Taro Toyoizumi1
1
Brain Science Institute, RIKEN, Wakoshi, 351-0198, Japan; 2Centre for
Neuroscience, University of Strathclyde, Glasgow, G4 ORE, UK
E-mail: erin.munro@brain.riken.jp
BMC Neuroscience 2015, 16(Suppl 1):P264
Brain states can be classified as synchronized (large amplitude low
frequency oscillations) or desynchronized (small amplitude high
frequency activity). [1] Synchronized states are marked by UP states/
phases characterized by global spiking and DOWN states/phases are
characterized by global silence in the cortex. In awake animals,
desynchronized states are associated with processing sensory input and
behavior while synchronized states are associated with quiet idling
conditions. During sleep, REM is considered desynchronized and slow-
wave sleep is considered synchronized. While desynchronized brain states
are often triggered by various kinds of neuronal input to cortical areas,
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
the exact mechanism at work during synchronized brain states is still
unclear. In particular, there are two hypothesized mechanisms for the
slow oscillation during slow-wave sleep: UP phases can be produced
either by traveling neocortical waves or a thalamo-cortical loop [2-4].
In our study, applying independent component analysis (ICA) to
recordings from rat neocortex reveals two different mechanisms during
synchronized activity. The mechanisms are distinguished by two key
neural sources identified by ICA: a strong broad source centered in layer
5 (BL5) and an apparent sub-cortical source producing clock-like
oscillations which resemble hippocampal theta oscillations (SUB). The
BL5-state often resembles cortically generated oscillations: UP phases are
initiated in deeper layers akin to traveling neocortical waves and the
oscillation is relatively slow. The SUB-state can resemble thalamo-
cortically generated oscillations: UP phases are initiated in layer 4 as well
as deeper layers and the oscillation is faster. These findings suggest that
both hypothesized mechanisms for the slow oscillation are at work in the
cortex - in alternation.
Acknowledgements: We would like to thank Austin Brockmeier for
helpful discussions with ICA. Waking data was originally published in
Sakata & Harris (2009) and Sakata & Harris (2012). This work was
supported by the RIKEN Brain Science Institute, Medical Research council
(MR/J004448/1), Tenovus Scotland (S11/1), Deafness Research UK (552:
STR:SS), Medical Research Council (MR/J004448/1), and Biotechnology and
Biological Sciences Research Council (BB/K016830/1).
References
1. Harris K, Thiele A: Cortical state and attention. Nat Rev Neur 2011,
12(9):509-523.
2. Massimini M, Huber R, Ferrarelli F, Hill S, Tononi G: The sleep slow
oscillation as a traveling wave. J Neurosci 2004, 24(31):6862-6870.
3. Crunelli V, Hughes SW: The slow (<1Hz) rhythm of non-REM sleep: a
dialogue between three cardinal oscillators. Nat Neurosci 2010, 13(1):9-17.
4. David F, Schmiedt JT, Taylor HL, Orban G, Di Giovanni G, Uebele VN,
Renger JJ, Lambert RC, Leresche N, Crunelli V: Essential thalamic
contribution to slow waves of natural sleep. J Neurosci 2013,
33(50):19599-19610.
P265
Self-organization of complex cortex-like wiring in a spiking neural
network model
Daniel Miner*, Jochen Triesch
Department of Neuroscience, Frankfurt Institute for Advanced Studies,
Frankfurt am Main, Hessen 60486, Germany
E-mail: miner@fias.uni-frankfurt.de
BMC Neuroscience 2015, 16(Suppl 1):P265
Understanding the structure and dynamics of cortical connectivity is vital
to understanding cortical function. Experimental data strongly suggest that
local recurrent connectivity in the cortex is significantly non-random,
exhibiting above-chance bidirectionality, an overrepresentation of certain
triangular motifs, and a heavy-tailed distribution of synaptic efficacies [1].
Additional evidence suggests a significant distance dependency to
connectivity over a local scale of a few hundred microns [2], and particular
patterns of synaptic turnover dynamics [3]. It is currently not understood
how many of these non-random features arise. Gaining understanding,
then, of the processes that lead to these complexities would provide
valuable insights into the development and computational functionality of
the cortex. While previous work has attempted to model some of the
individual features of local cortical wiring, there is no model that
comprehensively begins to account for all of them.
Here we present a spiking network model of a Layer V-like cortical slice
culture (panel B) that, via the interactions of a few simple biologically
motivated plasticity mechanisms, qualitatively reproduces many of these
non-random effects, such as synaptic weight (panel A) and triangular motif
distribution (panel C) Additionally, it reproduced experimentally observed
synaptic growth and efficacy dynamics [3]. These plasticity mechanisms
include spike timing dependent plasticity, synaptic normalization,
homeostatic firing threshold adaptation, pruning of zero-efficacy synaptic
connections, and the distance-dependent generation of new synaptic
connections. As a spiking, topographic extension to the previously
developed SORN family of models [4,5], there is also evidence suggesting
that these plasticity mechanisms endow recurrent networks with powerful
Page 173 of 200
learning abilities. Our model suggests that mechanisms of self-organization
arising from a small number of plasticity rules provide a parsimonious
explanation for numerous experimentally observed non-random features
of recurrent cortical wiring.
References
1. Song S, Sjstrm PJ, Reigl M, Nelson S, Chklovskii DB: Highly nonrandom
features of synaptic connectivity in local cortical circuits. PLoS Biol 2005,
3:e68.
2. Perin R, Berger TK, Markram H: A synaptic organizing principle for cortical
neuronal groups. Proc Natl Acad Sci U S A 2011, 108:5419-5424.
3. Yasumatsu N, Matsuzaki M, Miyazaki T, Noguchi J, Kasai H: Principles of
long-term dynamics of dendritic spines. J Neurosci 2008, 28:13592-13608.
4. Lazar A, Pipa G, Triesch J: SORN: a self-organizing recurrent neural
network. Front Comput Neurosci 2009, 3:23, October.
5. Zheng P, Dimitrakakis C, Triesch J: Network self-organization explains the
statistics and dynamics of synaptic connection strengths in cortex. PLoS
Comput Biol 2013, 9:e1002848.
P266
Key features of neural variability emerge from self-organized sequence
learning in a deterministic neural network
Christoph Hartmann1*, Andreea Lazar2, Jochen Triesch1
1
Frankfurt Institute for Advanced Studies (FIAS), Frankfurt, Germany; 2Ernst-
Strngmann Institute (ESI), Frankfurt, Germany
E-mail: chartmann@fias.uni-frankfurt.de
BMC Neuroscience 2015, 16(Suppl 1):P266
Cortical responses to identical stimuli show high trial-to-trial variability. This
variability is commonly interpreted as resulting from internal noise.
However, much of the variability can be explained by the pre-stimulus
spontaneous activity [1]. In fact, the contribution of this spontaneous
activity to the evoked response is sufficiently strong to bias perceptual
decisions [2]. Importantly, spontaneous activity is structurally similar to
evoked activity [3] and this similarity may be the result of learning an
internal model of the environment during development [4]. Consistent
with this idea, spontaneous activity seems to be a superset of possible
evoked responses [5] and trial-to-trial variability drops at stimulus onset [6].
At present, it is unclear how these features of neural variability arise in
cortical circuits.
Here, we show that all of these phenomena emerge in a completely
deterministic self-organizing recurrent network (SORN) model [7]. The
network consists of recurrently connected excitatory and inhibitory
populations of McCulloch-Pitts units. The dynamics are shaped by spike-
timing dependent plasticity (STDP) and homeostatic plasticity mechanisms
in response to structured input sequences. After a period of self-
organization, during which the network learns an internal model of the
input sequences, we observe all phenomena mentioned above: evoked
responses and perceptual decisions can be predicted from prior
spontaneous activity, spontaneous activity outlines the realm of evoked
responses, Fano factors drop at stimulus onset, and spontaneous activity
closely matches evoked activity patterns. In addition, the network
produces the common signs of Poissonian variability in single units.
In sum, our model demonstrates that key features of neural variability
emerge in a fully deterministic network from self-organized sequence
learning via the interaction of STDP and homeostatic plasticity mechanisms.
These results suggest that the high trial-to-trial variability of neural
responses need not be taken as evidence for noisy neural processing
elements.
References
1. Arieli A, Sterkin A, Grinvald A, Aertsen A: Dynamics of ongoing activity:
explanation of the large variability in evoked cortical responses. Science
1996, 273:1868-1871.
2. Hesselmann G, Kell C a, Eger E, Kleinschmidt A: Spontaneous local
variations in ongoing neural activity bias perceptual decisions. Proc Natl
Acad Sci U S A 2008, 105:10984-10989.
3. Kenet T, Bibitchkov D, Tsodyks M, Grinvald A, Arieli A: Spontaneously
emerging cortical representations of visual attributes. Nature 2003,
425:954-956.
4. Berkes P, Orbn G, Lengyel M, Fiser J: Spontaneous cortical activity reveals
hallmarks of an optimal internal model of the environment. Science (80- )
2011, 331:83-87.
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 174 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

Figure 1(abstract P265) Panel A: Mature synaptic weight distribution (in logarithmic space) for a simulation run with associated Gaussian
(lognormal in linear space) fit. Panel B: Topological graph of mature network for a simulation run. Panel C: Triangular motif count (relative to random
and corrected for an overrepresentation of bidirectional connections, similar to [1]) for topological graph of mature network. Motif key to right

5. Luczak A, Barth P, Harris KD: Spontaneous events outline the realm of possible 6. Churchland MM, et al: Stimulus onset quenches neural variability: a
sensory responses in neocortical populations. Neuron 2009, 62:413-425. widespread cortical phenomenon. Nat Neurosci 2010, 13:369-378.

Figure 1(abstract P266) Two example results after self-organization. a) The neural variability drops at stimulus onset. b) Spontaneous and evoked
activity become more similar during learning
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
7. Lazar A, Pipa G, Triesch J: Emerging Bayesian priors in a self-organizing
recurrent network. Artificial Neural Networks and Machine Learning - ICANN
2011, 127-134.
P267
Modulation of hippocampal gamma oscillations by acetylcholine:
insights from mathematical and in vitro optogenetic models
Ruth Betterton1*, Jack Mellor1, Krasimira Tsaneva-Atanasova2
1
School of Physiology and Pharmacology, University of Bristol, Bristol, BS8
1TD, UK; 2College of Engineering, Mathematics and Physical Sciences,
University of Exeter, Exeter, EX4 4QF, UK
E-mail: r.betterton@bristol.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P267
A neuronal oscillation involves the rhythmic, synchronous firing of a
population of cells. Oscillations found throughout the cortex can be
separated into bands of differing frequencies which are associated with
various behavioural states. Gamma oscillations (30 - 100 Hz) occur
coincidently with attention, sensory processing and learning and memory.
The hippocampus, known for its role in learning and memory, shows
gamma activity in vivo [1] and gamma oscillations can be induced in
in vitro slices [2]. The release of acetylcholine (ACh) correlates with
increases in oscillatory power in vivo [3] and knockout of specific ACh
receptor subtypes provides evidence for this scenario [4]. To further
investigate the role of ACh in the modulation of gamma oscillations we
have utilised both in vitro and computational techniques.
We implemented a mathematical model of the CA3 region of the
hippocampus based on [5]. Using Hodgkin-Huxley single compartmental
neurons, we verify that a network of 80 excitatory pyramidal cells and 20
inhibitory interneurons is able to produce oscillatory activity within the
gamma range.
We developed an optogenetic system (see [6,7]) to induce gamma
oscillations enabling us to test modulation by specific acetylcholine
receptors. Male mice received stereotaxic injection into the CA3 region of
the hippocampus of a viral vector (AAV5) containing channelrhodopsin
(hChR2(H134R)) under the control of the CaMKIIa promoter. Stimulation
of the ChR expressing CA3 pyramidal cell bodies with short light pulses
(5-50 ms) evoked action potentials and stimulation of Schaffer collateral
axons elicited robust synaptic responses in the CA3 and CA1 regions that
were blocked by the application of NBQX (10 M) or TTX (1 M). Local
field potential recordings showed that a 1s step optical stimulation
induced low power and low frequency gamma oscillations which
attenuated over time. In vivo, gamma oscillations are often found nested
within an overlying theta oscillation [8]. Correlating well with these in
vivo recordings, theta frequency (5 Hz) sine wave optical stimulation
induced higher power and higher frequency gamma oscillations with less
attenuation over a 1s period.
We introduced, a similar, 5 Hz sine depolarising input to the pyramidal
cells in our mathematical model and found that it was able to induce
oscillations at gamma frequency.
By manipulating specific currents within the model, we predicted the
effect of specific ACh receptor subtype activation on gamma oscillations.
These predictions were supported by our in vitro experimental evidence
showing that we found that activation of ACh receptors did indeed
modulate gamma oscillations with M1 receptor having a major effect.
References
1. Csicsvari J, Jamieson B, Wise KD, Buzski G: Mechanisms of gamma oscillations
in the hippocampus of the behaving rat. Neuron 2003, 37:311-322.
2. Plhalmi J, Paulsen O, Freund TF, Hjos N: Distinct properties of carbachol-
and DHPG-induced network oscillations in hippocampal slices.
Neuropharmacology 2004, 47:381-389.
3. Morrosu F, Portas C, Mascia MS, Casu MA, F M, Giagheddu M, Imperato A,
Gessa GL: Microdialysis measurement of cortical and hippocampal
acetylcholine release during sleep-wake cycle in freely moving cats.
Brain research 1995, 671:329-332.
4. Fisahn A, Pike FG, Buhl E, Paulsen O: Cholinergic induction of network
oscillations at 40Hz in the hippocampus in vitro. Nature 1998, 394:186-188.
5. Kopell N, Brgers C, Pervouchine D, Malerba P, Tort A: Gamma and Theta
Rhythms in Biophysical Models of Hippocampal Circuits. Hippocampal Microcircuits,
Springer Series in Computational Neuroscience 5 New York: Springer 2010.
Page 175 of 200
6. Akam T, Oren I, Mantoan L, Ferenczi E, Kullmann DM: Oscillatory dynamics
in the hippocampus support dentate gyrus-CA3 coupling. Nature
Neuroscience 2012, 15(5):763-768.
7. Pastoll H, Solanka L, van Rossum MCW, Nolan MF: Feedback Inhibition
Enables Theta-Nested Gamma Oscillations and Grid Firing Fields. Neuron
2012, 77:141-154.
8. Colgin LL, Denninger T, Fyhn M, Hafting T, Bonnevie T, Jensen O, Moser M,
Moser EI: Frequency of gamma oscillations routes flow of information in
the hippocampus. Nature 2009, 462:353-357.
P268
A spatiotemporal model of spine calcium dynamics in the hippocampus
Thom Griffith1*, Jack Mellor2, Krasi Tsaneva-Atanasova3
1
Department of Engineering Maths, University of Bristol, Bristol, UK; 2School
of Physiology and Pharmacology, University of Bristol, Bristol, UK;
3
Department of Mathematics, University of Exeter, Exeter, UK
E-mail: thom.griffith@bristol.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P268
Ca 2+ -signalling in dendritic spines is required for NMDA receptor-
dependent synaptic plasticity at glutamatergic synapses in the
hippocampus [1]. However, it is not clear whether plasticity induction is
dependent solely on the global signal, i.e., the spine volume-averaged Ca2+
signal; or whether plasticity induction is also sensitive to Ca2+-channel
nanodomain signaling [2]. A working hypothesis of this work is that
temporal and spatial variations in postsynaptic intracellular [Ca2+]-fields
may be significant factors governing the signalling cascades that lead to
either long-term synaptic potentiation or depression. Direct measurement
of [Ca2+] distributions in dendritic spines is experimentally difficult but we
can investigate this hypothesis using mathematical models of Ca 2+
diffusion.
We have developed a spatio-temporal model of Ca 2+ diffusion in three
dimensions. We then study our model using finite element methods. The
model allows predictions of intracellular [Ca 2+ ]-field responses to
combinations of pre- and post-synaptic spikes with nanometre and
millisecond spatio-temporal resolution. Our results so far indicate that Ca2+
signalling is highly spatially non-uniform and that Ca2+ signal differences
between induction protocols is dependent on location within the spine. This
has implications for the ultimate biological role of the Ca2+ signal given that
the relevant receptors in the spine are organised inhomogeneously [3].
Acknowledgements: Support for this work was provided by the EPSRC,
UK (EP/I013717/1).
References
1. Malenka RC, Bear MF: LTP and LTD: an embarrassment of riches. Neuron
2004, 44(1):5-21.
2. Chen Y, Sabatini BL: Signaling in dendritic spines and spine
microdomains. Current Opinion in Neurobiology 2012, 22(3):389-396.
3. Mori MX, Erickson MG, Yue DT: Functional stoichiometry and local
enrichment of calmodulin interacting with Ca2+ channels. Science 2004,
304(5669):432-435.
P269
Transient synchrony in delayed coupled neuronal networks
Zahra G Esfahani1*, Alireza Valizadeh1,2
1
Department of Physics, Institute for Advanced Studies in Basic Sciences,
Zanjan, Iran; 2School of Cognitive Sciences, IPM, Niavaran, Tehran, Iran
E-mail: nzahra-ghasemiiasbs@ac.ir
BMC Neuroscience 2015, 16(Suppl 1):P269
In this study, we propose that in a pool of neurons recurrently coupled
through delayed synaptic connections transient patterns of synchrony can
be observed due to the changing incoming stimuli, in continuance of some
recent works [1]. Transient synchrony between spiking activity of the
neurons has been reported in different sensory tasks e.g. visual and
olfactory system [2,3].
We have shown that the critical role of the delay is to prepare
connections that their synchronizing/desynchronizing effect changes
when they receive different levels of stimuli [4,5]. In a suitable range of
parameters, need not to be fine-tuned, an initially incoherent firing of the
neurons can turn to coherent network oscillation when the mean input is
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P269) Raster Plot of an all-to-all network of 200 homogenously coupled neurons with time dependent stimuli (the violet
curve), which switches between asynchronous incoherent state to a synchronous state when the input is changed. All the neurons are excitatory
and external input to each neuron comprises a constant current chosen from a narrow normal distribution and an independent Gaussian white noise.
The blue diagram presents the network activity
changed -not necessarily increasedthrough sensory or control input
(Figure 1). It is important to note that such an ability of the network to
select frequencies of the oscillation is based on the presence of the delay
in communication between neurons. In a network in which the
components communicate instantaneouslywith delays ignoredthe
neurons either spike synchronously or asynchronously depending on the
connections properties and regardless of the value of the input current
and the frequency of the spiking of the neurons.
Conclusion: We have shown that the ability of a neural network to switch
between coherent and incoherent firing, may be dependent on the delay
in communication between neurons. It has been shown that two
reciprocally coupled neurons can fire inphase if the delays lie in the region
where the phase response curve of the neurons have negative slope,
otherwise their firing is antiphase. In the larger networks where the
neurons connect to several other neurons, inphase firing state remains
stable where instead of antiphase state, several stable states appear. This is
related to geometric frustration in condensed matter physics where a
plenitude of distinct ground states are ensued by the lattice structure as in
Ising system.
References
1. Gollo LLeonardo, Breakspear Michael: The frustrated brain: from dynamics
on motifs to communities and networks. Philo. Trans. of the Royal Society
B: Biol. Sci 2014, 369(1653):20130532.
2. Maxim Bazhenov, et al: Model of transient oscillatory synchronization in
the locust antennal lobe. Neuron 2001, 30(2):553-567.
3. Tatsuya Mima, et al: Transient interhemispheric neuronal synchrony
correlates with object recognition. The Journal of Neuroscience 2001,
21(11):3942-3948.
4. Sadjad Sadeghi, Valizadeh Alireza: Synchronization of delayed coupled
neurons in presence of inhomogeneity. Journal of computational
neuroscience 2014, 36(1):55-66.
5. Esfahani GhZahra, Valizadeh Alireza: Zero-Lag Synchronization Despite
Inhomogeneities in a Relay System. PloS one 2014, 9(12):e11268.
P270
Stabilizing synchrony with heterogeneity
Ehsan Bolhasani1,2*, Alireza Valizadeh1,2
1 systems 2002, 2:3-54.
Department of Physics, Institute for Advanced Studies in Basic Sciences,
Zanjan, Iran; 2School of Cognitive Sciences, Institute for Studies in Theoretical
Physics and Mathematics, Niavaran, Tehran, Iran
E-mail: ebhasani@iasbs.ac.ir
BMC Neuroscience 2015, 16(Suppl 1):P270
Correlation and synchrony between spike trains of neurons can arise from
shared input they receive from other neurons, or from direct connections
between the neurons. Physiological heterogeneity can destabilize both
Page 176 of 200
coupling-induced and shared input-induced synchronization [1]. In the
classical models of synchronization, collective state of a system of
coupled oscillators is determined by outcome of rivalry between
synchronizing effect of connections and desynchronizing effect of
inhomogeneity [2]. Yet there are examples of the systems in which
inhomogeneity enhances synchrony [3,4]. In this study we have
developed a general framework for the correlation of coupled neuronal
oscillators with a given phase sensitivity and we have shown that in
some cases, for identical neurons, synchronized state is an unstable
attractor and arbitrarily weak noise can destroy synchrony.
Inhomogeneity in such systems could stabilize synchrony by providing an
asymmetric basin of attraction around the stable phase-locked state. This
in turn results in a sharper PDF for the time difference between spikes of
the two neurons in presence of noise (see Figure 1A). The analytic results
are obtained by solving Fokker-Planck equations for phase oscillators and
it is also shown that in presence of stochastic inputs, the most probable
phase difference between spike times of the two neurons does not
coincide with the stable point of the deterministic equations. Numerical
tests on LIF neurons confirm the positive role of inhomogeneity in
stabilizing synchrony and increasing correlation of spike trains. The only
difference is that for the model neurons with biologically realistic phase
response curve (PRC), the time difference between the spikes of two
neurons in the stable state increases with inhomogeneity, in the case of
LIF neurons, they lock in almost zero phase lag for sufficiently small
values of inhomogeneity (see Figure 1B).
Conclusion: While usually physiological heterogeneity is considered as a
destructive factor for the synchrony of spike trains, we have shown that
correlation of spike trains of two coupled neurons can be enhanced by
the mismatch of intrinsic firing rates of the neurons. The result, obtained
from analytic solution of pulse-coupled phase oscillators, is valid for all
type-I neuronal oscillators with strictly positive phase reset curves,
including LIF neurons.
References
1. Kopell N, Ermentrout B: Mechanisms of phase-locking and frequency
control in pairs of coupled neural oscillators. Handbook of dynamical
2. Kuramoto Y: Self-entrainment of a population of coupled nonlinear
oscillators. In international symposium on mathematical problems in
theoretical physics Springer Berlin Heidelberg 1975, 420-422.
3. Braiman Y, Ditto WL, Wiesenfeld K, Spano ML: Disorder-enhanced
synchronization. Physics Letters A 1995, 206(1):54-60.
4. Bolhasani E, Azizi Y, Valizadeh A: Direct connections assist neurons to
detect correlation in small amplitude noises. Frontiers in computational
neuroscience 2013, 7.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P270) (A) PDF for the phase difference of two type1 phase oscillators in the presence of noise for three levels of frequency
mismatch. Theoretical predictions (solid lines) are in accordance with numerical results shown as histograms. Dash lines show the phase-locked states in
the absence of noise. (B) Cross-Correlogram for two levels of current mismatch for LIF neuron model. Inset: the magnitude of maximum correlation for
two neurons shows the positive effect of mismatch in enhancement of correlation
P271
Novel modes in a Wilson-Cowan network
Jeremy Neuman1*, Jack D Cowan2, Wim van Drongelen3
1
Dept. of Physics, University of Chicago, Chicago, IL 60637, USA; 2Dept. of
Mathematics, University of Chicago, Chicago, IL 60637, USA; 3Dept. of
Pediatrics, University of Chicago, Chicago, IL 60637, USA
E-mail: jneuman@uchicago.edu
BMC Neuroscience 2015, 16(Suppl 1):P271
Spontaneous [1] and synaptically-driven neural activity exhibit a wide
variety of dynamics. In the latter case, recent experiments using spike-
triggered LFPs [2] have been able to classify stimulated behavior into two
distinct categories: 1) traveling waves with smooth attenuation when the
input is weak; and, 2) localized responses when the impulse is strong.
Unfortunately, our knowledge of the mechanisms behind these
differences is lacking on both the cellular and network scales.
This study, employing the spatiotemporal mean-field Wilson-Cowan
equations [3], provides a model for the nature of these two modes at the
population level. Just as in [2], we detect damped traveling waves with
exponential decay when the input is relatively small. When the stimulus
increases, the activity stays localized as evidenced by the large slope in
the peaks of the activity.
Conclusions: Understanding the contrast between traveling and localized
activity in synaptically-driven neural networks is an important task due to
its wide range of applications to areas such as vision, sleep and epilepsy.
Here, it was shown that a simple population model exhibits both modes
of behavior. This result allows us to further study how various network
properties such as density and connectivity strength give rise to these
kinds of activity.
Acknowledgement: This work was supported by the Dr. Ralph and
Marian Falk Medical Research Trust Fund.
References
1. Mao B, Aronov D, Froemke R, Yuste R: Dynamics of Spontaneous Activity
in Neocortical Slices. Neuron 2001, 32(6):883-898.
2. Nauhaus I, Busse L, Ringach D, Carandini M: Robustness of Traveling
Waves in Ongoing Activity of Visual Cortex. J. Neuroscience 2012,
32(9):3088-3094.
3. Wilson H, Cowan J: A mathematical theory of the functional dynamics
of cortical and thalamic nervous tissue. Kybernetik 1973, 13(2):55-80.
Page 177 of 200
P272
Proof of concept: a spatial modular small-world self-organises by
adaptive rewiring
Nick Jarman1,2*, Chris Trengove1, Erik Steur1, Ivan Tyukin2,3,
Cees van Leeuwen2
1
Perceptual Dynamics Laboratory, University of Leuven, Leuven, Flemish
Brabant, B3000, Belgium; 2Department of Mathematics, University of
Leicester, Leicester LE1 7RH, UK; 3Saint-Petersburg State Electrotechnical
University, Saint-Petersburg, Saint Petersburg 197376, Russia
E-mail: nick.jarman@ppw.kuleuven.be
BMC Neuroscience 2015, 16(Suppl 1):P272
A small-world network is a network that reconciles two opposing
properties, segregation and integration. It is this reconciliation that gives
rise to the impressive information processing capacity of the human brain;
segregation provides a platform for information processing, whilst
integration provides for the fast transmission of information. However, the
connectivity structure of the brain is not static [1]; it changes on multiple
time-scales; on a relatively fast time-scale, synaptic plasticity takes place,
whilst on a slower time-scale there is rewiring of brain connectivity
through growth of axons and dendrites. This structural plasticity depends
on the even faster time-scale of neural activity. But the relationship is
symbiotic: patterns of synchronous activity are, of necessity, mediated by
the brain connectivity structure. Gong & van Leeuwen [2] showed that
rewiring of an initially random network - adaptive rewiring - in a model of
spontaneous cortical activity gives rise to a particular type of network
connectivity structure: a modular small-world. In order to improve the
applicability of such a model to the cortex, spatial characteristics of cortical
connectivity need to be respected. For this purpose we consider networks
endowed with a metric by embedding them into a physical space. Such
spatial constraints may represent wiring and metabolic costs in the brain.
We provide an adaptive rewiring model with a spatial distance function
and a corresponding spatially local rewiring bias [3].
Conclusion: The resulting rewiring scenarios showed a spatial layout of
the connectivity structure, in which topologically segregated modules
correspond to spatially segregated regions, and these regions are linked
by long-range connections (see Figure 1, A and B). Greater realism and
increased efficiency and robustness of the symbiosis of activity and
structure is achieved compared to non-spatial adaptive rewiring. Thus, the
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 178 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

Figure 1(abstract P271) Simulations showing damped traveling waves and a localized response when the input strength is varied. (A-B)
Diagrams of total activity at equally-spaced times. (C-D) Plots of distance traveled for the peaks (as percentage of global maximum) of curves in (A-B),
respectively. (C) is fitted to an exponential as in [2] while (D) fits a linear curve with slope that can be approximated as infinity on the given scale. (E-F)
Plots of the rate of propagation of the peaks in (A-B), respectively. Both fits are linear but the localized behavior moves at a much smaller speed. Insets
are reproduced from [2]

Figure 1(abstract P272) A, Network adjacency matrix organised to optimise visual presentation of modular structure. B, Units on the sphere
colour-coded to identify distinct modules

principle of locally biased adaptive rewiring may explain both the
topological connectivity structure and spatial distribution of connections
between neuronal units in a large-scale cortical architecture.
References
1. Zhang LI, Poo MM: Electrical activity and development of neural circuits.
Nat Neurosci 2001, 4:1207-1214.
2. Gong P, van Leeuwen C: Evolution to a small-world network with chaotic
units. Europhys Lett 2004, 67:328-333.
3. Jarman N, Trengove C, Steur E, Tyukin I, v. Leeuwen C: Spatially
constrained adaptive rewiring in cortical networks creates spatially
modular small world architectures. Cognitive Neurodynamics 2014,
8:479-497.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
P273
Lateral connections synchronize population activity in a spiking neural
network model of midbrain superior colliculus
Bahadir Kasap*, John van Opstal
Donders Institute for Brain, Cognition and Behaviour; Dept. Biophysics,
Radboud University Nijmegen, Nijmegen, the Netherlands
E-mail: b.kasap@donders.ru.nl
BMC Neuroscience 2015, 16(Suppl 1):P273
Saccades are rapid and ballistic eye-head gaze shifts between points of
interest in the visual field. They are crucial for gathering high-resolution
visual information. The midbrain superior colliculus (SC) generates
saccadic eye-movement commands for downstream oculomotor circuits.
It contains an eye-centered, gaze-motor map that relates the location of a
Gaussian-shaped neural population to the intended movement vector.
The gaze-motor map mediates the spatiotemporal transformation for eye-
head orienting gaze shifts to peripheral targets [1]. Electrophysiological
recordings have shown that SC neurons exhibit some remarkable activity
properties that depend on both their anatomical position and the
resulting saccade trajectory [2].
Here, we propose a biologically plausible spiking neural network model
that is constrained by the observed firing patterns of real SC neurons for
visually evoked saccades. The functional two-dimensional network model
reproduces the spike trains of single neurons in recorded SC populations
for saccades with different amplitudes and directions.
The network model consists of a 2D grid of neurons, representing the
gaze-motor map. The adaptive integrate-and-fire neurons [3] portray the
observed site-dependent bursting profiles of individual SC neurons
through distinct intrinsic biophysical properties, whereas Mexican-hat
shaped lateral connections ensure the observed synchronized population
activity by a soft winner-takes-all mechanism.
We argue that our model offers a basis for neuronal algorithms of
spatiotemporal transformations and bio-inspired optimal control signal
generators.
Acknowledgements: This work is funded by the European Commission
through FP7 Marie Curie ITN project NETT (Grant nr. 289146) (BK) and
the Radboud University Nijmegen (AvO).
References
1. Ottes FP, van Gisbergen JAM, Eggermont JJ: Visuomotor fields of the
superior colliculus: a quantitative model. Vision Research 1986,
26(6):857-873.
2. Goossens HHLM, van Opstal AJ: Optimal control of saccades by spatial-
temporal activity patterns in the monkey superior colliculus. PLoS
Computational Biology 2012, 8(5):e1002508.
3. Brette R, Gerstner W: Adaptive exponential integrate-and-fire model as an
effective description of neuronal activity. Journal of Neurophysiology 2005,
94(5):3637-3642.
P274
Channel-specific input/output transformations arising from the
interaction between dynamic synapses and subthreshold oscillations
Roberto Latorre1, Joaquin J Torres2, Pablo Varona1*
1
Dpto. de Ingeniera Informtica, Escuela Politcnica Superior, Universidad
Autnoma de Madrid, 28049 Madrid, Spain; 2Dpto. de Electromagnetismo y
Fsica de la Materia, and Institute Carlos I for Theoretical and Computational
Physics, University of Granada, Granada, Spain
E-mail: pablo.varona@uam.es
BMC Neuroscience 2015, 16(Suppl 1):P274
Subthreshold oscillations are observed in a wide variety of neurons in the
nervous system (e.g. see [1,2]). They typically appear associated to
resonance phenomena that allow, for instance, the implementation of
intrinsic memory mechanisms for the detection of specific spike sequences
in single neurons and neural networks [3-5]. On the other hand, synaptic
transmission and, therefore, the corresponding input/output transformation
in the postsynaptic cell are affected by recent presynaptic activity. Dynamic
synapses modulate incoming information increasing (facilitation) or
decreasing (depression) the postsynaptic response. This modulation
provides additional characteristic time scales to the single neuron and
network processing [6,7]. In the context of a neuron displaying subthreshold
Page 179 of 200
oscillations, the interplay between these additional time scales and the
characteristic intrinsic fast and slow dynamics underlying the neurons
oscillatory activity can significantly affect its input/output transformation.
Neuronal intrinsic subthreshold oscillations and dynamic synapses underlie
several computational properties both at the single neuron and the
network levels. Traditionally, intrinsic oscillations and dynamic synapses
have been studied separately and their interaction has attracted almost no
attention. In this work, we use a conductance-based neuron model and a
dynamic synapse to investigate how intrinsic subthreshold oscillations and
the short-term plasticity of a depressing synapse act together to shape its
resonant properties and the corresponding input/output transformation.
Our results suggest that factors such as the maximum hyperpolarization
level, the oscillation amplitude and frequency or the resulting firing
threshold can be modulated by synaptic depression. This shapes the
postsynaptic neurons resonant properties arising from the subthreshold
oscillation and leads to complex channel-specific input/output relations.
Thus, a low-cost modification in synaptic parameters can produce a
significant different response. This complex synaptic-dependent input/
output transformation allows the implementation of cost-effective
information discrimination mechanisms in single neurons by just tuning
the depression level of the synaptic channel without modifying the
intrinsic neural dynamics.
Acknowledgements: Authors acknowledge support by MINECO TIN2012-
30883 and FIS2013-43201-P.
References
1. Llins R, Yarom Y: Oscillatory properties of guinea-pig inferior olivary
neurones and their pharmacological modulation: an in vitro study. J
Physiol 1986, 376:163-182.
2. Giocomo LM, Zilli EA, Fransn E, Hasselmo ME: Temporal Frequency of
Subthreshold Oscillations Scales with Entorhinal Grid Cell Field Spacing.
Science (80- ) 2007, 315:1719-1722.
3. Baroni F, Varona P: Subthreshold oscillations and neuronal input-output
relationships. Neurocomputing 2007, 70:1611-1614.
4. Baroni F, Torres JJ, Varona P: History-dependent excitability as a single-
cell substrate of transient memory for information discrimination. PLoS
One 2010, 5:e15023.
5. Latorre R, Aguirre C, Rabinovich MI, Varona P: Transient dynamics and
rhythm coordination of inferior olive spatio-temporal patterns. Front
Neural Circuits 2013, 7:138.
6. Klug A, Borst JGG, Carlson BA, Kopp-Scheinpflug C, Klyachko VA, Xu-
Friedman MA: How Do Short-Term Changes at Synapses Fine-Tune
Information Processing? J Neurosci 2012, 32:14058-14063.
7. Torres JJ, Kappen HJ: Emerging phenomena in neural networks with
dynamic synapses and their computational implications. Front Comp
Neurosci 2013, 7.
P275
Regularization of a half-center oscillator network by closed-loop control
Irene Elices*, Pablo Varona
Grupo de Neurocomputacin Biolgica, Departamento de Ingeniera
Informtica, Escuela Politcnica Superior, Universidad Autnoma de Madrid,
Madrid, 28049, Spain
E-mail: irenelices@uam.es
BMC Neuroscience 2015, 16(Suppl 1):P275
Central Pattern Generators (CPGs) are neural circuits that control muscle
functioning by means of rhythmic patterns. These networks are usually
built up on a minimal configuration based on reciprocal inhibitory
connections responsible for the production of alternating spiking-bursting
activity. Experimental observations in the crustacean pyloric CPG show that
most neurons, when isolated, present a highly irregular, in fact chaotic,
bursting activity [1-3]. This rich intrinsic dynamics provides flexibility for
negotiating rhythms through the reciprocal inhibitory connections
between neurons which lead to the regularization of the chaotic behavior
when the neurons interact within the circuit [4].
Closed-loop interactions are typically used in electrophysiological
experiments using dynamic clamp protocols [5] and have been
generalized for different description levels of the nervous system [6].
In this work we show that feedback protocols can also be used in
theoretical studies to search for specific dynamics or explore the
parameter space of a given model. We have built a CPG model based on a
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P275) Panel A: Activity of the system using the closed-loop algorithm. Panel B: Schematic representation of the closed-loop
protocol
minimal network with two neurons connected by bidirectional fast
chemical inhibitory synapses. The network generates alternating bursting
activity in the neurons, which can be regular or irregular depending on the
maximal conductances of the inhibitory synapses. We employ a simple
adaptive closed-loop protocol to regularize the alternating chaotic activity
of the model (see Figure 1 panel A). This protocol adapts online the
maximal conductance of one of the synapsis to achieve the aimed regular
alternating bursting activity (see Figure 1, panel B). Moreover, the closed-
loop algorithm can be used to automatically map the region of maximal
conductance values that lead to regular activity.
Acknowledgements: Authors acknowledge support by MINECO TIN2012-
30883 and ONRG grant N62909-14-1-N279.
References
1. Abarbanel HD, Huerta R, Rabinovich MI, Rulkov NF, Rowa PF, Selverston AI:
Synchronized action of synaptically coupled chaotic model neurons.
Neural Comput 1996, 8(8):1567-1602.
2. Elson RC, Huerta R, Abarbanel HD, Rabinovich MI, Selverston AI: Dynamic
control of irregular bursting in an identified neuron of an oscillatory
circuit. J Neurophysiol 1999, 82(1):115-122.
3. Varona P, Torres JJ, Abarbanel HD, Rabinovich MI, Elson RC: Dynamics of
two electrically coupled chaotic neurons: experimental observations and
model analysis. Biological Cybernetics 2001, 84(2):91-101.
4. Selverston AI, Rabinovich MI, Abarbanel HD, Elson R, Szcs A, Pinto RD,
Huerta R, Varona P: Reliable circuits from irregular neurons: a dynamical
approach to understanding central pattern generators. J Physiology-Paris
2000, 94(5):357-374.
5. Destexhe A, Bal T (Eds): Dynamic-Clamp: From Principles to Applications
Springer, New York 2009.
6. Chamorro P, Muiz C, Levi R, Arroyo D, Rodrguez FB, Varona P:
Generalization of the dynamic clamp concept in neurophysiology and
behavior. PLoS One 2012, 7:e40887.
P276
Two-channel models of medial and superior olive based on
psychoacoustics
Jaroslav Bouse1*, Vaclav Vencovsky1,2
1 1. Wang D, Brown GJ: Computational Auditory Scene Analysis: Principles,
Department of Radioelectronics, Faculty of Electrical Engineering, Czech
Technical University in Prague, Prague, 162 27, Czech Republic; 2Musical
Acoustics Research Center, Academy of Performing Arts in Prague, Prague,
118 00, Czech Republic
E-mail: bousejar@fel.cvut.cz
BMC Neuroscience 2015, 16(Suppl 1):P276
Two nuclei in each hemisphere, the medial superior olive (MSO) and the
lateral superior olive (LSO), are responsible for decoding interaural time
difference (ITD) and interaural level difference (ILD), respectively [1]. This
allows us to localize sound sources in the horizontal plane.
Neurons in the MSO are usually modeled as coincidence detectors [1].
Two main approaches to model the MSO appeared during last decades.
The most utilized approach (introduced by Jeffress [2]) uses a delay line
Page 180 of 200
together with coincidence detectors. The model assumes that the
individual neurons in the MSO are sensitive to the different ITD. Another
approach (introduced by von Bksy [3]) accounts for two channels in
the brain competing with each other. The concept of two channel model
was further developed by Dietz [4] (called rate-code model), and Pulkki
and Hirvonen [5] (called count-comparison model), who incorporated
new physiological findings. The LSO neurons are consistently reported as
an Excitation-Inhibition type [1] and mainly ILD sensitive, thus it can be
modeled by a simple subtracter.
We present phenomenological models mimicking the functions of the
MSO and LSO. The models incorporate the latest physiological findings.
However, the models were not designed to simulate responses of the
neurons in the MSO and LSO. Instead, the models give data directly
comparable with results of subjective lateralization experiments. To
simplify the readability, we call these models according to the nuclei
which they are mimicking, the MSO and the LSO model. As a front-end, a
model of peripheral ear (taken from the literature) is utilized. The MSO
model is Excitation-Inhibition type, in principle similar to the two
channels, rate-code or the count-comparison models. The LSO model is
Excitation-Inhibition type with a simple subtracting unit simulating the
process inside the LSO. There is one MSO and LSO model for each
hemisphere. Outputs from both hemispheres are compared following the
two channels principle. Interaction between the MSO and the LSO model
output is not yet possible in this implementation.
We compare the outputs of the designed models with the results of
psychoacoustic experiments showing the lateralization of pure tones and
narrow band noises. The data illustrating the lateralization of pure tones
with ITD or ILD was reproduced from the literature [6]. A psychoacoustic
experiment was conducted to measure the subjective lateralization of a
narrowband noise (central frequency 380 and 760 Hz, bandwidth 1 ERB)
with interaural phase or level differences. Seven and five normal hearing
listeners participated in the case of interaural phase and level differences
test, respectively. A good agreement between the model data and
subjective lateralization was observed.
Acknowledgements: This work was supported by the Grant Agency of the
Czech Technical University in Prague, grant No. SGS14/204/OHK3/3T/13.
References
Algorithms, and Applications Wiley-IEEE Press 2006.
2. Jeffress LA: A place theory of sound localization. J Comp Physiol Psychol
1948, 41(1):35-39.
3. von Bksy G: Zur Theorie des Hrens. ber das Richtungshren bei
einer Zeitdifferenz oder Lautstrkeungleichheit der beiderseitigen
Schalleinwirkungen. Psysik Zeitschr 1930, 824-835, 857-868.
4. Dietz M, Ewert SD, Hohmann V, Kollmeier B: Coding of temporally
fluctuating interaural timing disparities in a binaural processing model
based on phase differences. Brain Res 2008, 1220:234-245.
5. Pulkki V, Hirvonen T: Functional count-comparison model for binaural
decoding. Acta Acust united Ac 2009, 95(5):883-900.
6. Yost WA: Lateral position of sinusoids presented with interaural intensive
and temporal differences. J Acoust Soc Am 1982, 70(2):337-409, 1981.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
P277
A self-organizing neural network for neuromuscular control
Praveen Shankar1*, Sharmila Venugopal2
1
Department of Mechanical and Aerospace Engineering, California State
University Long Beach, CA, USA; 2Department of Integrative Biology and
Physiology, University of California Los Angeles, CA, USA
E-mail: vsharmila@ucla.edu
BMC Neuroscience 2015, 16(Suppl 1):P277
Adaptive technology holds great promise for sensorimotor rehabilitation
in people suffering from spinal cord injury, neuromuscular disease and
stroke. With a long-term goal of developing adaptive technology for
diagnosis and rehabilitation of neuromuscular dysfunction, we begin the
development of a self-organizing neural network (SNN) that compensates
for reduced neural drive. We suggest that the self-organizing architecture
that adds or deletes nodes online to generate suitable compensatory
muscle excitation (Figure 1A) is an apt mechanism to emulate the motor
pool behavior of recruitment and de-recruitment of motor units during
muscle force generation. Using a virtual muscle [1] resembling the
Figure 1(abstract P277) A. Schematic showing the virtual muscle-
SNN system; F1, F2, .. Fn are radial basis functions and w1, w2, ..
wn are weights for summation. B. Simulation of normal (Slow-Fast
motor unit ratio - 2:4), abnormal (Slow-Fast motor unit ratio - 3:3)
muscle force and, compensation by SNN
Page 181 of 200
human biceps brachii, we demonstrate the augmentation of neural
excitation by the SNN to compensate for abnormal muscle force due to
change in the number of motor units.
Reference
1. Cheng E, Brown I, Loeb G: Virtual muscle: a computational approach to
understanding the effects of muscle properties on motor control. Journal
of Neuroscience Methods 2000, 101:117-130.
P278
Developing and validating an isotrigon texture discrimination task
using Amazon Mechanical Turk
John WG Seamons1*, Marconi S Barbosa1, Jonathan D Victor2,
Dominique Coy1, Ted Maddess1
1
Eccles Institute for Neuroscience, John Curtin School of Medical Research,
ANU, Canberra, ACT 0200, Australia; 2Department of Neurology & Neuroscience,
Weill Cornell Medical College, 1300 York Ave, New York 10021, USA
E-mail: john.seamons@anu.edu.au
BMC Neuroscience 2015, 16(Suppl 1):P278
The human visual system must employ mechanisms to minimize
informational redundancy whilst maintaining that which is behaviorally
relevant [1,2]. Previous research has concentrated on two-point
correlations via spatial frequency and orientation tuning. Higher-order
correlations are less studied, but they may inform us about cortical
functioning [3]. Isotrigon textures can be used to probe the sensitivity of
the human visual system as their structure is exclusively due to 4th and
higher-order spatial correlations [4]. Although artificially generated, the
same features that give isotrigons salience also create salience in natural
images [2]. We implemented an isotrigon discrimination task using the
crowdsourcing platform Amazon Mechanical Turk (mTurk) [5]. An
important secondary aim was to evaluate the suitability of mTurk for visual
psychometric studies as very few exist [6].
960 HITs were uploaded to mTurk and 121 nave subjects participated.
Based on data quality, 91% of HITs were retained at a cost of $0.132 AUD
per HIT. The mTurk data was compared to two supervised lab datasets. Lab
and mTurk performance functions were very similar (Figure 1A) and highly
correlated (Figure 1B). Bland-Altman plots were examined and the mean
lab/mTurk coefficient of repeatability was 15.5%. Factor analysis was
performed on the combined data and 2 principal factors were identified.
Previous studies support that the number of mechanisms is less than
10 [7] and more likely 2-4 [8,9]. The congruence between the lab and
mTurk data is striking considering the unsupervised mode of delivery. In
conclusion, mTurk is an underutilized platform for visual psychometric
research which can produce data of comparable quality to lab samples at
reduced cost and increased scale.
References
1. H Barlow: Redundancy reduction revisited. Network 2001, 12(3):241-253.
2. G Tkacik, JS Prentice, JD Victor, V Balasubramanian: Local statistics in
natural scenes predict the saliency of synthetic textures. Proc Natl Acad
Sci U S A 2010, 107(42):18149-18154.
3. JD Victor: Isodipole Textures: A Window on Cortical Mechanisms of Form
Processing. Early Vision and Beyond MIT Press: Papathomas TVC, C.; Gorea,
A; Kowler, E 1995, 99-107.
4. T Maddess, Y Nagai, AC James, A Ankiewcz: Binary and ternary textures
containing higher-order spatial correlations. Vision Res 2004,
44(11):1093-1113.
5. W Mason, S Suri: Conducting behavioral research on Amazons
Mechanical Turk. Behavior research methods 2012, 44(1):1-23.
6. J Freeman, CM Ziemba, DJ Heeger, EP Simoncelli, JA Movshon: A
functional and perceptual signature of the second visual area in
primates. Nature neuroscience 2013, 16(7):974-981.
7. RR Taylor, T Maddess, Y Nagai: Spatial biases and computational
constraints on the encoding of complex local image structure. J Vis 2008,
8(7):19 11-13.
8. T Maddess, Y Nagai: Discriminating isotrigon textures [corrected]. Vision
Res 2001, 41(28):3837-3860.
9. T Maddess, Y Nagai, JD Victor, RR Taylor: Multilevel isotrigon textures.
J Opt Soc Am A Opt Image Sci Vis 2007, 24(2):278-293.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P278) 1A: Color map of median texture discrimination performance versus isotrigon texture type. Lab datasets DC (84 HITs)
and M1-6 (270 HITs). mTurk datasets Live1 (480 HITs) and Live2 (480 HITs). 1B: Pearsons correlation coefficients between datasets (abbreviations as above)
P279
Neural model for multi-stability in visual action recognition
MA Giese1*, L Fedorov1, R Vogels2
1
Department of Cognitive Neurology, Section Computational Sensomotorics,
CIN & HIH, University Clinic Tbingen, Germany; 2Laboratorium voor Neuro-
en Psychofysiologie, KU Leuven, Leuven, Belgium
E-mail: martin.giese@uni-tuebingen.de
BMC Neuroscience 2015, 16(Suppl 1):P279
Introduction: The visual perception of body movements shows interesting
dynamical properties. Biological motion perception can show perceptual
multi-stability with respect to the perceived walking direction [1]. In addition,
body motion perception is subject to adaptation, as is demonstrated by the
existence of high-level after-effects and of fMRI repetition suppression in
relevant areas. Existing neural models for action recognition do not account
for these phenomena. We present neurodynamical model that reproduces
these phenomena, and which allows to study the interplay between multi-
stability, adaptation and intrinsic fluctuations in body motion perception.
Methods: The core of the model is a two-dimensional neural field [2],
whose first dimension encodes the time-course of the action stimulus (in
terms of snapshot sequences) while the second dimension encodes the
stimulus view. In addition, inspired by results on adaptation processes in
area IT [3], the model includes two different types of adaptation processes,
modeling firing rate fatigue and input fatigue, and a Gaussian noise process.
The model is fitted to psychophysical data on multi-stability in action
perception, and the details of the adaptation processes were fitted,
exploiting electrophysiological results from area IT. The multi-stability of the
2D neural field model can be mathematically analyzed based on a novel
approach that is based on level-sets [4].
Results / Discussion: A version of the model for the representation of
static stimuli reproduces quantitatively electrophysiological results on
Page 182 of 200
neural adaptation in area IT [3]. The model accounts in a unifying way for
multiple neurodynamic phenomena in action recognition: (i) temporal
sequence-selectivity for movies with different sequential orders of the
stimulus frames; (ii) perceptual multi-stability for silhouette stimuli that are
compatible with multiple views; (iii) sizes of the electrophysiologically
observed adaptation effects for the repeated presentation of the same
action stimulus [5,6]. A detailed quantitative analysis shows that perceptual
switches in the neural dynamics are primarily induced by fluctuations
(noise), not by neural adaptation. In addition, the model predicts a new
highly efficient stimulus for the demonstration of adaptation effects in
action perception, which induces stronger adaptation within the neural
representation than the previously applied stimuli. Finally, it turns out that
adaptation effects induced by action stimuli are critically dependent on
the neural mechanism of adaptation. Firing-rate fatigue mechanisms seem
to result in much smaller adaptation effects for dynamic stimuli than
mechanisms based on input fatigue, when the adaptation effects for static
stimuli are matched.
Acknowledgements: Supported by EC FP7 ABC PITN-GA-011-290011,
HBP FP7-604102, Koroibot FP7-611909, COGIMON H2020-644727, DFG GI
305/4-1, DFG GZ: KA 1258/15-1, and BMBF, FKZ: 01GQ1002A.
References
1. Vanrie J, Dekeyser M., Verfaillie K: Bistability and biasing effects in the
perception of ambiguous point-light walkers. Perception 2004,
33(5):547-560.
2. Amari S: Dynamics of pattern formation in lateral-inhibition type neural
fields. Biol Cybern 1977, 27(2):77-87.
3. de Baene W, Vogels R.: Effects of adaptation on the stimulus selectivity of
macaque inferior temporal spiking activity and local field potentials.
Cereb Cortex 2010, 20(9):2145-2165.
4. Coombes S, Schmidt H, Bojak I: Interface dynamics in planar neural field
models. J. Math. Neurosci 2012, 2:9.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
5. Caggiano V, Pomper JK, Fleischer F, Fogassi L, Giese M, Thier P.: Mirror
neurons in monkey area F5 do not adapt to the observation of
repeated actions. Nat Commun 2013, 4:1433.
6. Kilner JM, Kraskov A, Lemon RN: Do monkey F5 mirror neurons show
changes in firing rate during repeated observation of natural actions?
J Neurophysiol 2013, 111(6):1214-1226.
P280
Optimization of input parameters to a CN neuron model to simulate its
activity during and between epileptic absence seizures
Parimala Alva1*, Lieke Kros2, Oscar H.J Eelkman Rooda2, Chris I De Zeeuw3,
Rod Adams1, Neil Davey1, Freek E Hoebeek2, Volker Steuber1
1
Science and Technology Research Institute, University of Hertfordshire,
Hatfield AL10 9AB, UK; 2Department of Neuroscience, Erasmus Medical
Center, Rotterdam, the Netherlands; 3Netherlands Institute for Neuroscience,
Royal Dutch Academy for Arts and Sciences, Amsterdam, the Netherlands
E-mail: p.alva2@herts.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P280
Absence seizures are often attributed to the synchronized oscillatory
activity in the thalamo-cortical regions of the brain, and they can be
detected by the presence of spike-wave-discharges (SWDs) in the
electroencephalogram (EEG). The cerebellar nuclei (CN), which have
afferent connections to these regions, may also play a role in the
propagation of these seizures. Some CN neurons in Cacna1atottering (tg)
mice phase-lock their spiking activity to the peaks of the SWDs in the EEG
during the absence seizures. These CN neurons are deemed to
participate in the absence seizures. To investigate if certain types of CN
neuron are more likely to participate, we performed Growing Neural Gas
(GNG) clustering [1], using different subsets of the feature set that
consisted of CV, mean Cv2, log CV, log-interval entropy, permutation
entropy, firing rate, burst index, pause ratio, burst-like spike ratio, mean ISI,
mode ISI, median ISI, min ISI of the interictal parts of the spike-trains. The
subset of the feature set that produced the best separation of clusters was
CV, burst-like spike ratio, and mode ISI (Figure 1).
We then used a morphologically realistic conductance-based model of an
excitatory CN projection neuron [2] to simulate the interictal activity of a
Figure 1(abstract P280) 2D projection, using principal component
analysis, of the clusters formed as a result of GNG clustering of
the CN neurons interictal activity using CV, burst-like spike ratio
and mode ISI. (+) indicate cells that participate in the seizure and (o)
indicate the cells that do not participate based on the measures, FFT
based Z-score and modulation frequency. The black (+) indicate the
selected participating cluster for the EA based optimization. The grey
dots indicate the output data points for every 5th generation of the EA
and the arrows show the progression of the output data points from
the initial position indicated by 0 to the interictal cluster center
Page 183 of 200
participating CN neuron. We selected a cluster where all neurons
participated in the seizures, indicated by black crosses (+) in Fig. 1, and
applied an Evolutionary algorithm (EA) to optimize excitatory and
inhibitory input parameters to the CN neuron model such as spike rates,
noise, burst parameters, synchronicity, synaptic weights so that the
output data point of the EA moved closer to the centre of the selected
cluster. The results of the EA indicated that the CN neurons that
participated in absence seizures received either a synchronous, bursting
inhibitory input or a synchronous, bursting excitatory input. Next, we
modified the EA such that the initial input parameters of the CN neuron
model resulted in an output data point nearest to the center of the
selected interictal cluster, and ran the optimization to move the output
data point to the center of the cluster that was formed from the ictal
counter-parts of the CN neurons of the selected participating cluster.
Surprisingly, a very small change in input parameters could result in a
shift from the interictal to the ictal cluster centre and result in a transition
to CN neuron activity as observed during seizures. However, when we
blocked the Purkinje cell input to the CN neuron model by maintaining
the inhibitory synaptic weight at zero, the output data point never
reached the center of the ictal cluster. This suggests that blocking
Purkinje cell input to the CN neuron can prevent the CN neuron from
participating in the absence seizure.
References
1. Fritzke B: A Growing Neural Gas Network Learns Topologies. Neural
Information Processing Systems 1994, 7:625-632.
2. Steuber V, Schultheiss N, Silver RA, Schutter E, Jaeger D: Determinants of
synaptic integration and heterogeneity in rebound firing explored with
data-driven models of deep cerebellar nucleus cells. J Comp Neurosci
2011, 30(3):633-658.
P281
A spiking neuron network model for the delayed motion direction
discrimination task
Liu-Tao Yu1, Si Wu2, Da-Hui Wang1,2*
1
School of Systems Science, Beijing Normal University, Beijing 100875, China;
2
State Key Laboratory of Cognitive Neuroscience & Learning, Beijing Normal
University, Beijing 100875, China
E-mail: wangdh@bnu.edu.cn
BMC Neuroscience 2015, 16(Suppl 1):P281
Drawing the correct decision in a given situation usually depends on past
experiences. While the neural basis of decision making per se has been
studied extensively, how a memory of a previous situation could influence
a later decision is less clear. One behavioral paradigm for investigating
decision making in relation to an earlier reference is the delayed motion
direction discrimination task (DMD) [1]. In this task, a variable reference
direction is first presented to the subject by displaying a coherent random
dot kinematograms (RDK). After memorizing, a delayed presentation of the
test direction (RDK) follows and the subject needs to report whether the
test direction is clockwise (CW) or anticlockwise (ACW) rotated in relation to
the reference direction (Figure 1A). Although phenomenological models
have been proposed, the underlying neural mechanism remains relatively
unknown. Here, we hypothesize that a mechanism based on the idea of
asymmetric connections similar to those used in neural networks
explaining angular path integration in the head-direction system [2], might
explain the behavioral results.
To test this hypothesis, we build a spiking neuron network model and
investigate whether it can solve the DMD task. The network consists of
three parts (Figure 1.B): 1) a working memory circuit (WMC) 2) two
information extraction circuits, referred to as clockwise-preferred circuit
(CPC) and anticlockwise-preferred circuit (APC) respectively, and 3) a
decision-making circuit. At the core of the network is the assumption of
an asymmetric offset and rotational invariance of the connectivity profile.
The former ensures that CPC (or APC) has stronger response when the
test direction is CW (or ACW) rotated with respect to the reference
direction and the latter guarantees the feasibility of the network under
variable references. The simulations demonstrate that the proposed
network is indeed capable of solving the DMD task (Figure 1.C) showing
a similar detection threshold (2) as previous behavioral data [3].
Furthermore, the model correctly predicts that with higher similarity of
reference and test direction or lower coherence level of the RDK, the
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P281) A. DMD task: either ACW or CW has to be reported in relation to the reference. B. Network structure. Pyramid cells of
WMC (CPC and APC) are arranged on a circle according to their preferred motion directions and with rotation invariant Gaussian shaped lateral
connections. Note that from WMC to CPC and APC connectivity is offset by -45/+45. Right box indicates the decision making network. C. Network
activity during the DMD task. Reference (90) is applied to the WMC (blue line), while the test (100) is applied to CPC and APC (green line). Due to the
connectivity structure (see B), the group A of the decision circuit activates stronger than group C indicating a correct ACW decision
performance gets worse (lower accuracy and longer reaction time). Our
results suggest a possible neural mechanism for memory-guided decision
making.
References
1. Watamaniuk SN, Sekuler R: Temporal and spatial integration in dynamic
random-dot stimuli. Vision Res 1992, 32:2341-7.
2. Song P, Wang X-J: Angular path integration by moving hill of activity": a
spiking neuron model without recurrent excitation of the head-direction
system. J Neurosci 2005, 25:1002-14.
3. Sekuler R, Watamaniuk SNJ, Blake R: Stevens Handbook of Experimental
Psychology Wiley, New York: Yantis, S. & Pashler, H. 2002, 1:121-176.
P282
Prediction of ion channel parameter differences between groups of
young and aged pyramidal neurons using multi-stage compartmental
model optimization
Timothy Rumbell1*, Danel Dragulj2, Jennifer Luebke3, Patrick Hof1,
Christina M Weaver2
1
Fishberg Department of Neuroscience and Friedman Brain Institute, Icahn
School of Medicine at Mount Sinai, New York, NY 10029, USA; 2Department
of Mathematics, Franklin and Marshall College, Lancaster, PA 17604, USA;
3 fitting these four neurons reveal, for example, that in the aged neuron
Department of Anatomy and Neurobiology, Boston University School of
Medicine, Boston, MA 02118, USA
E-mail: timothy.rumbell@mssm.edu
BMC Neuroscience 2015, 16(Suppl 1):P282
Electrophysiological recording and morphological reconstruction of
pyramidal neurons from layer 3 prefrontal cortex of young and aged
adult rhesus monkeys reveals higher firing rates and structural
differences in aged neurons relative to young [1]. Prior computational
modeling of these neurons has demonstrated that morphology alone is
insufficient to account for functional differences, predicting that
passive and active biophysical parameters differ between age groups
[2]. Here we use compartmental models featuring several independent
ion channels to provide insight into the precise electrophysiological
parameters that underlie excitability differences between these
neurons.
We fit ion channel parameters in a neuron model using a rapid, multi-
stage, semi-automated approach based on our previous optimization
protocol [3], resulting in a population of models representing the target
neuron. First, we use a stereotypical pyramidal cell morphology [4] with
size scaled to match the target neuron surface area. Second, we use
differential evolution (DE) to tune passive and H-current parameters that
shape the models response to subthreshold somatic current injection.
Third, we add 7 voltage- and calcium-dependent ion channels with free
maximal conductance and kinetic parameters, and set up an optimization
that will fit model output to empirical responses to current injections that
Page 184 of 200
evoke repetitive action potential firing. We manually select error functions
for the suprathreshold optimization, then use a Latin hypercube sampling
(LHS) to generate a space-filling set of parameter combinations. We then
simulate each point in the LHS, and use the results to calculate weights
for error functions. This novel weighting method is fully automated, and
scales error functions based on their expected contribution to the total
error during the suprathreshold optimization. Fourth, we perform
suprathreshold optimization using DE, with the free parameters and
weighted error functions chosen in stage three. The result of this
optimization is a population of models that fit the target neuron. DE
allows for fully parallel optimization, so we generate 256 well-fitting
models within 80 hours of optimization through the Neuroscience
Gateway [5].
We have used this method to generate populations of models for 1
aged and 3 young neurons. Linear discriminant analysis reveals that
these four populations are readily distinguished within the parameter
space. Models of the three young neurons have different ion channel
parameters despite similarities in electrophysiological responses.
Principal component analysis (PCA) reveals that all models of the aged
neuron can be separated from models of the young neurons along the
first principal component. Preliminary differences between parameters
the L-type calcium channel activates more slowly and has greater
maximal conductance, and voltage dependence of the persistent
sodium channel is lower. Generation of populations of models for
other empirically characterized young and aged neurons is underway
with our optimization protocol. Analytical techniques such as PCA will
help generate predictions about intracellular changes during normal
aging.
References
1. Chang YM, Rosene DL, Killiany RJ, Mangiamele LA, Luebke JI: Increased
action potential firing rates of layer 2/3 pyramidal cells in the prefrontal
cortex are significantly related to cognitive performance in aged
monkeys. Cereb Cortex 2005, 15(4):409-418.
2. Coskren P, Luebke JI, Kabaso D, Wearne SL, Yadav A, Rumbell T, Hof PR,
Weaver CM: Functional consquences of age-related morphologic
changes to pyramidal neurons of the rhesus monkey prefrontal cortex. J
Comput Neurosci 2014, Dec 20 [Epub ahead of print].
3. Rumbell T, Dragulji D, Luebke J, Hof P, Weaver CM: Automatic fitness
function selection for compartmental model optimization. BMC
Neuroscience 2014, 15(Suppl 1):O5.
4. Traub RD, Gloveli T, Whittington MA: Fast rhythmic bursting can be
induced in layer 2/3 cortical neurons by enhancing persistent Na+
conductance or by blocking BK channels. J Neurophysiol 2003, 89:909-921.
5. Sivagnanam S, Majumdar A, Yoshimoto K, Astakhov V, Bandrowski A,
Martone ME, Carnevale NT: Introducing the Neuroscience Gateway. IWSG,
CEUR Workshop Proceedings 2013 [CEUR-WS.org].
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
P283
What must come down goes up - the effect of noise on weights in
spike-timing-dependent plasticity
Michael Klein*, Angelo Cangelosi, Thomas Wennekers
Centre of Robotics and Neural Systems, Plymouth University, PL4 8AA,
Plymouth, UK
E-mail: michael.q.klein@gmail.com
BMC Neuroscience 2015, 16(Suppl 1):P283
In large-scale spiking neural models that learn with spike-timing-
dependent plasticity (STDP), it is a crucial, but difficult problem to balance
synaptic potentiation (necessary to learn a task) and synaptic depression
(necessary to counteract accidental weight increase and to balance overall
firing rates). Adding random noise and choosing parameters such that
inhibition slightly dominates excitation [1,2] is considered one way of
accomplishing this. With the learning rule being standard STDP [2], the
parameter that determines whether inhibition or excitation dominates is
the so-called a-parameter, defined as the ratio between parameters that
determine amount of depression and those that determine amount of
potentiation [3]. If the a-parameter is set to a value greater than 1, the
weights are assumed to go down. However, using numerical simulations,
we demonstrate that this is not the case. Using standard leaky integrate-
and-fire neurons, we ran 10 simulations per parameter set (250s each),
with a-parameters A = {0.5, 0.9, 1.0, 1.1, 1.25, 1.5} and noise frequencies F
= {5, 10, 20, 40, 80, 160}. We also varied the maximum weight and the
initial weight of the synapse from 0 to 1 in small intervals. In contrast to
our assumptions, we find a weight increase with a-parameters greater
than 1 (see Figure 1). The complex weight dynamics observed can be
explained by the interplay of 3 factors. 1. The a-parameter: if it is greater
than 1 it tends to drive the weights down, given that firing is uncorrelated.
2. The impact of the pre-synaptic spike on the post-synaptic potential
(determined by synaptic weight, capacitance of the neuron, or time
constant of the synapse): if the impact is strong enough pre-synaptic
spikes will cause post-synaptic spikes and firing will no longer be
uncorrelated. Hence, weights will tend to go up (for example, if the initial
values of the weights are too high). 3. Noise frequency: There is a positive
interaction between noise frequency and a-parameter in driving weights
down. If the frequency is high enough, the a-parameter can be low. If the
frequency is low the a-parameter needs to be very high. High frequency
drives the weights down, because the post-synaptic neuron operates in
input averaging mode, firing more regular than the pre-synaptic spikes
that evoked it, and, therefore, being decorrelated from it [2]. These
findings give insights in how to set parameters (in particular a-parameter,
initial weights and noise frequency) to achieve a desired weight dynamics
in large-scale STDP models.
Acknowledgements: This work was supported under EPSRC Grant EP/
J004561/1 (BABEL).
Figure 1(abstract P283) A plot of the weight strength after training
for different frequencies with an initial weight of 0.5. As expected,
the weight increases for a-parameters smaller than 1. However,
the weight also increases for a-parameters larger than 1 for
frequencies below 80Hz.
Page 185 of 200
References
1. Gerstner W, Kistler WM: Spiking Neuron Models Cambridge University Press,
New York 2002.
2. Song S, Miller KD, Abbott LF: Competitive Hebbian learning through
spike-timing dependent synaptic plasticity. Nature Neuroscience 2000,
3:919-926.
3. Gerstner W, Kistler WM, Naud R, Paninski L: Neural Dynamics - From single
neurons to networks and models of Cognition. Cambridge University Press
2014, July.
P284
A Hebbian cell assembly based neural field model for the remote
associate task and creative search
Ivana Kajic*, Thomas Wennekers
School of Computing and Mathematics, Plymouth University, Plymouth,
Devon PL48AA, UK
E-mail: twennekers@plymouth.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P284
The Remote Associates Test (RAT) is widely used in experimental
psychology to assess verbal creative thinking [1]. Subjects are given triplets
of words (the cues, e.g., spoon, coin, and quick) and instructed to find
a fourth word that relates to all the cues (the target, e.g., silver). Cues
presented separately usually elicit strong associations with numerous other
words (close associates), but since their combination triggers the specific
target word, it is said that this target is remotely associated. Several
studies have addressed the cognitive search process in RAT, which, given
sufficient search time, can go through a number of candidate words
before the target is actually found (e.g., [2]).
In order to elucidate possible neural mechanisms underlying the RAT, we
here propose a neural field model that combines the idea of Hebbian cell
assemblies (CAs) with a winner-take-all competitive process for assembly
selection (WTA), and inhibition-of-return (IOR) [3] to allow for multiply-
constrained autonomous searches on word-networks. Words are represented
by localised populations of neurons (assemblies) mapped over the 2D model
sheet; they can be activated by inputs representing the cues. Semantic
similarity is implemented by the lateral network of hetero-associative
connections in such a way that strongly related words have stronger mutual,
possibly asymmetric, connections. Local excitatory connections support the
amplification and latching of activity in an assembly. Local inhibition, evoked
by activity in an assembly, can overcome the excitation of that assembly
after some time, switch it off, and provide a lasting inhibition-of-return. This
allows for the sampling of other potential target words. Global inhibition
(WTA) furthermore keeps the total activity in the network small, such that at
any time only a small number of CAs can be active.
We present computer simulations of the search process in a space of
associated words under different conditions regarding association
strength between close and remote associates, search time limits, and
the impact of noise on RAT performance. We also analyse the network
dynamics to determine conditions that allow for the existence of remote
associates and that support a Markovian search dynamic as described
experimentally [2].
Acknowledgements: This work has been supported by the Marie Curie
Initial Training Network FP7-PEOPLE-2013-ITN, CogNovo, grant number
604764, http://cognovo.eu.
References
1. Mednick SS: The associative basis of the creative process. Psychological
Review 1962, 69:220-232.
2. Smith KA, Huber DE, Vul E: Multiply-constrained semantic search in the
Remote Associates Test. Cognition 2013, 128(1):64-75.
3. Itti L, Koch C: Computational Modelling of Visual Attention. Nat Rev
Neurosci 2001, 2(3):194-203.
P285
Recurrent networks expect
Adam Ponzi*, Jeffery R Wickens
Okinawa Institute of Science and Technology Graduate University (OIST),
Onna, Okinawa 904 0495, Japan
E-mail: adamp@oist.jp
BMC Neuroscience 2015, 16(Suppl 1):P285
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P285) Excess correct classification probability of
on-time target stimuli compared to mistimed target stimuli (100
ms early, black ; 100 ms late, red) in a temporally regular
streaming task with inter-trial-interval 800 ms. Results are plotted
versus the Lyapunov exponent of the autonomous network dynamics
How the brain anticipates future events is one of the most intriguing
questions in neural information processing. The brain does not simply
respond to the external world but in predictably structured environments
like natural language, birdsong and somatosensory flow in fluids predicts it
[1]. Reactions are improved by predictive encoding of what sensory stimuli
may occur as well as when they may occur. Quantitative studies of implicit
timing in streaming perceptual discrimination tasks show that
performance is enhanced if stimuli occur at expected times according to
an established rhythm. The neural basis for implicit timing is not fully
understood but oscillatory entrainment mechanisms have been suggested
[2]. Recent studies confirm that low-frequency brain oscillations do
become phase entrained in such tasks. Phase entrainment increases with
the temporal regularity of the sensory stream and correlates with
enhanced discrimination performance [3]. It is not limited to sensory
cortices; larger scale cortical networks, as well as subcortical networks like
the basal ganglia [4], may also be coherently modulated by the
predictability of stimuli streams. The presence of large scale oscillatory
entrainment suggests that network mediated neuronal ensemble
dynamics may be involved. Recurrent neural networks generate complex
but reproducible temporally extended dynamical activity patterns in
response to input stimuli [5,6]. Such transient activity patterns have been
suggested to provide a natural substrate for working memory and a
representation of elapsed time.
Here we add to the understanding of how random recurrent neural
networks support neural information processing by demonstrating that
temporal expectation also naturally emerges from their dynamics. We
show that the weakly chaotic oscillations generated by recurrent networks
can be phase synchronized [7] by temporally regular stimulus sequences.
We show network responses are maximally discriminative when stimuli fall
at their preferred phase in phase entrained networks (Figure 1).
Discriminability increases continuously with both temporal regularity and
stimulus type predictability which also interact. Our results do not depend
on specific network characteristics, are resilient to the presence of network
noise and random distractor stimuli and are consistent with multiple
streaming perceptual discrimination studies.
References
1. Nobre AC, Correa A, Coull JT: The hazards of time. Curr. Op. Neurobio 2007,
17(4):465-470.
2. Large EW, Jones MR: The dynamics of attending: How people track time-
varying events. Psych. Rev 1999, 106(1):119.
Page 186 of 200
3. Cravo AM, Rohenkohl G, Wyart V, Nobre AC: Temporal expectation
enhances contrast sensitivity by phase entrainment of low-frequency
oscillations in visual cortex. J. Neurosci 2013, 33(9):4002-4010.
4. Geiser E, Notter M, Gabrieli JD: A corticostriatal neural system enhances
auditory perception through temporal context processing. J. Neurosci
2012, 32(18):, 6177.
5. Buonomano DV, Merzenich MM: Temporal information transformed into a
spatial code by a neural network with realistic properties. Science 1995,
267:1028-1030.
6. Ponzi A, Wickens JR: Optimal Balance of the Striatal Medium Spiny
Neuron Network. PLoS comp bio 2013, 9(4):e1002954.
7. Pikovsky A, Zaks M, Rosenblum M, Osipov G, Kurths J: Phase
synchronization of chaotic oscillations in terms of periodic orbits. Chaos
1997, 7(4):680-687.
P286
A novel method for spatial source localization using ECoG and SEEG
recordings in human epilepsy patients
Chaitanya Chintaluri*, Daniel K Wjcik
Department of Neurophysiology, Nencki Institute of Experimental Biology,
Warsaw - 02093, Poland
E-mail: c.chintaluri@nencki.gov.pl
BMC Neuroscience 2015, 16(Suppl 1):P286
In recent decades, there has been increasing interest in surgical treatments
of patients with pharmacologically intractable epileptic seizures. In some
cases, noninvasive methods do not sufficiently localize epileptogenic foci
and invasive methods of presurgical evaluation are necessary. In some
patients, craniotomy is required, where-in electrodes are placed directly
over the cortex - these recordings are referred to as electrocorticography
(ECoG). Another invasive method is based on intra-cortical depth
electrodes which are stereotactically steered into deep cerebral structure -
these are referred to as stereoencephalography (SEEG). Based on these
recordings of electrical potentials at corresponding electrode positions, the
spatial location of the sources of epileptic activity in the brain, that is to be
lesioned, needs to be estimated. Improving the precision of locating these
sources - referred to as source imaging, derived from ECoG and SEEG
recordings remains a critical goal in the field [1].
Based on our previous studies [2], we propose a new method - kESI
(kernel Electrical Source Imaging), which takes into account realistic brain
morphology and spatial variations in brain conductivity. kESI can localize
multiple sources, and is flexible to arbitrary electrode positions. Therefore
this method can be effectively used for a specific patients case. The core
of kESI is in the construction of kernel functions requiring computation of
the potentials generated in the brain by numerous basis functions
covering the probed volume.
To test the robustness of the method, we generated ground truth data
using a simplified spherical brain model - with uniform conductivity,
and placed a dipole source inside it (Figure 1A). The electrodes were
placed on the surface of the sphere, and were also placed inside the
spherical volume emulating ECoG and SEEG style recordings
respectively. The potentials generated at these electrodes were
computed using Finite Element Methods (FEM) in FEniCS software, the
mesh was generated in gmsh. In kESI, this FEM model was used to
compute the potentials generated by the basis functions, and hence
obtain the reconstructed sources. To evaluate the accuracy, the root
mean squared difference (RMS) values between the ground-truth and
the reconstructed sources from kESI was computed in the orthogonal
planes passing through the dipole (Figure 1B). This was found to be
0.158 when using 652 electrodes (ECoG & SEEG), and 0.248 when using
100 electrodes (only ECoG) (Figure 1C).
Acknowledgements: Research funded from grants FP7-PEOPLE-2010-ITN
264872 NAMASEN, 2948/7.PR/2013/2.
References
1. K Kaiboriboon, HO Lders, M Hamaneh, J Turnbull, SD Lhatoo: EEG source
imaging in epilepsy - practicalities and pitfalls. Nat. Rev. Neurol 2012,
8:498-507.
2. J Potworowski, W Jakuczun, S ski, D Wjcik: Kernel Current Source
Density Method. Neural Comput 2012, 24:541-575.
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P286) A) Shows the VTK rendering of the simplified 3D brain model, the red-blue spheres indicate the placement of a dipole
inside the brain. B) Shows the two orthogonal planes passing through the dipole (red-blue spheres) and origin C) Reconstruction of sources from
simulated ECoG recordings, left:True CSD, right:kESI method (along XY Plane). The RMS difference between True CSD and kESI method, evaluated along
XY and XZ planes is 0.248
P287
Stimulus discrimination and association with Hebbian cell assemblies
Johannes M Auth1,2*, Timo Nachstedt1,2, Christian Tetzlaff1,2,
Florentin Wrgtter1,2
1
Third Institute of Physics, Georg-August-University, Gttingen, 37077,
Germany; 2Bernstein Center for Computational Neuroscience, Gttingen,
37077, Germany
E-mail: jauthphys@uni-goettingen.de
BMC Neuroscience 2015, 16(Suppl 1):P287
Learning or the formation of memories is a key component of biological
intelligence. One distinguishes between different types of memory as, for
instance, the declarative memory denoting the storage of facts and
concepts which are extracted from perceived stimuli. Thereby, an
important question is how the brain distinguishes between stimuli that
should be associated and those to be discriminated. Investigating this
mechanism, in psychophysical experiments, rats were exposed to gradually
morphed environments while recording activity in hippocampal place cells
[1,2]. Stability of the recorded place-dependent activation patterns was
dependent on the order in which the different environments were
explored. In other words, stimulus discrimination or association depends
on the order of presentation. Theoretical models of memory rely on the
hypothesis that memorization is implemented through activity-dependent
synaptic plasticity. Environmental patterns are reflected by correlated
activity in subgroups of neurons leading to high recurrent synaptic
efficacies among them. These groups of highly interconnected neurons are
named cell assemblies (CAs). This long-standing concept with its basic
ability of pattern completion is a promising approach to explain several
cognitive phenomena [3]. Combining the theory of CAs with the principle
of self-organizing maps (SOMs) [4], we present a biologically plausible
network model able to reproduce the upper experimental findings.
Basically our model consists of three populations of neurons. An input
population projects through random excitatory connections on a recurrent
layer. Neurons within the layer interact by excitatory synapses. Competition
is introduced by an inhibitory population mutually interconnected with
every layer neuron. For all neurons, we use a rate-coded neuron model with
sigmoidal transfer functions. The plasticity model combines (Hebbian)
synaptic plasticity and synaptic scaling (SPASS) [5]. Stimuli are inserted as
activity patterns in the input population passing them on via afferent plastic
synapses. Recurrent lateral connections in the layer reach a circular
neighborhood of fixed range and are also plastic. The incoming excitatory
synapses the inhibitory population receives as well as its outgoing
connections are non-plastic. In agreement with SOM, our concept maps
sufficiently differing stimuli to different representations (CAs) in the layer.
Comparable to the experimental procedures described above, we define
two non-overlapping activity patterns A and B. According with [1], first
we transform stimulus A gradually into B with a short pause between
Page 187 of 200
each step of the transition sequence. After A lead to the formation of
one memory representation, or rather a single CA, this CA is maintained
throughout the following learning protocol. Secondly, the protocol of [2]
demands a random presentation of the transition patterns. In this case,
we observe the formation of multiple representations. Thus, our model
reproduces the experimental findings described above. In contrast to
existing theoretical studies [6], it does not require stimulus-dependent
modifications of the plasticity rule itself. It further allows for investigating
the influence of different variations in the learning protocol as, for
instance, of the stimulus presentation time.
References
1. Wills TJ, Lever C, Cacucci F, Burgess N, OKeefe J: Attractor dynamics in the
hippocampal representation of the local environment. Science 2005,
308:873-876.
2. Leutgeb JK, Leutgeb S, Treves A, Meyer R, Barnes CA, McNaughton BL,
Moser MB, Mose EI: Progressive transformation of hippocampal neuronal
representations in morphed environments. Neuron 2005, 48:345-358.
3. Pulvermueller F, Garagnani M, Wennekers T: Thinking in circuits: toward
neurobiological explanation in cognitive neuroscience. Biol Cybern 2014,
108:573-593.
4. Kohonen T: Essentials of the self-organizing map. Neural Networks 2013,
37:52-65.
5. Tetzlaff C, Kolodziejski C, Timme M, Tsodyks M, Wrgtter F: Synaptic
Scaling Enables Dynamically Distinct Short- and Long-Term Memory
Formation. PLoS Comput Biol 2013, 9(10):e1003307.
6. Blumenfeld B, Preminger S, Sagi D, Tsodyks M: Dynamics of Memory
Representations in Networks with Novelty-Facilitated Synaptic Plasticity.
Neuron 2006, 52:383-394.
P288
Anticipative tracking in two-dimensional continuous attractor neural
networks
Yuanyuan Mi1,2*, Yan Xia1, Qi Gao1, Si Wu1
1
State Key Laboratory of Cognitive Neuroscience & Learning, Beijing Normal
University, Beijing 100875, China; 2Department of Neurobiology, Weizmann
Institute of Science, Rehovot 76100, Israel
E-mail: miyuanyuan@0102163.com
BMC Neuroscience 2015, 16(Suppl 1):P288
Time delays exist pervasively in neural information processing. The brain
needs to compensate for these delays in order to extract motion
information in time [1]. Experimental data reveal that in tracking moving
stimuli, neural systems generate anticipative responses to compensate for
delays [2], but how does the brain achieves this goal remains poorly
understood.
Continuous attractor neural networks (CANNs) are widely used as a
canonic model to describe the encoding of continuous features, such as
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 188 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

head-direction, moving direction, orientation or spatial location of an Table 1(abstract P290) Docking of Ligands on lower
object, in the brain. A CANN holds a continuous family of localized (THR361) binding sites of 5-HT3A
stationary states, called bumps. These bumps form a sub-manifold in the
network state space, on which the system is neutrally stable, and this Compound - (THR361 binding site) Binding energy (kcal/mol)
neutral stability endows the network with the capacity of tracking Cyclohexane -4.5
moving inputs smoothly [3]. Although a CANN is able to track a moving
input, its reaction time is always delayed with respect to the instant Eucalyptol -4.5
position of the input, due to that neurons responding to external input Menthol -5.6
and neuronal interaction via recurrent synapses consume time.
In this study, by incorporating spike frequency adaptation (SFA), a Propofol -6.0
negative feedback mechanism, in the network dynamics, we find that a
CANN is able to track a moving input anticipatively. We consider a two-
dimensional CANN, mimicking the neural circuit of place cells [4]. With Table 2(abstract P290) Docking of Ligands on upper
theoretical analysis and numerical simulation, we systematically explore (TYR346) binding sites of 5-HT3A
the dynamics of a 2D CANN, and find that: 1) in the absence of SFA or in
Compound - Rigid Docking - Flexible Docking -
the presence of weak SFA, the CANN holds static bump states; 2) when
(TYR346 binding Binding energy Binding energy
SFA is sufficiently strong, the CANN generates a traveling wave, in which
site) (kcal/mol) (kcal/mol)
a bump moves spontaneously in the network without relying on external
drive. The speed of the traveling wave is fully determined by the network 9- -7.0 -9.2
parameters, which we call the intrinsic speed vint of the network; 3) in tetrahydrocannabinol
response to a moving input with speed vext, the interplay between the (THC)
intrinsic mobility of the network and the speed of the external drive Cannabidiol (CBD) -7.0 -8.8
determines the network tracking performance, that is, a) when vext> vint,
the network state lags behind the external input; and b) when vext< vint,
the network state leads the external input. Interestingly, we find that the
anticipative time of the network, given by s/v ext with s the leading
distance, is approximately a constant for a wide range of input speed.
This property justifies the experimental finding that in a given brain area,
the anticipative time is approximately a constant independent of input
speed. By constructing a hierarchical model with multiple CANNs, our
model also reproduces the experimental finding that along the signal
pathway, the anticipative time of neurons increases.
Our study reveals that a neural system can utilize the intrinsic dynamics
of a neural circuit to implement anticipative responses to moving inputs.
It sheds light on our understanding of how the brain processes motional
information in a timely manner.
References
1. Nijhawan R: Visual prediction: psychophysics and neurophysiology of
compensation for time delays. Behav. Brain Sci 2008, 31:179-198.
2. Taube JS: The head direction signal: origins and sensory-motor
integration. Annu. Rev. Neurosci 2007, 30:181C207.
3. Fung C, Wong KYM, Wu S: A moving bump in a continuous manifold: a
comprehensive study of the tracking dynamics of continuous attractor
neural networks. Neural Comput 2010, 22:752-792.
4. Samsonovich A, McNaughton B: Path integration and cognitive mapping Figure 1(abstract P290) Experimental comparison [3] of the effects
in a continuous attractor neural network model. J. Neurosci 1997, of compounds structurally related to menthol on the 5-HT3
17:5900-5920. receptors

P290
The effect of 9-tetrahydrocannabinol, cannabidiol, menthol and
propofol on 5-hydroxytryptamine type 3 eeceptorsa computational
approach
Andreas C Schilbach*, Tatiana Prytkova, Susan Keun-Hang Yang
Schmid College of Science and Technology, Chapman University, Orange
CA, 92866, USA
E-mail: schil105@mail.chapman.edu
BMC Neuroscience 2015, 16(Suppl 1):P290
This study investigates the function of 5-HT type 3 (5-HT3) receptors
using a computational approach. Antagonists of the 5-HT3 receptor are
currently one of the most effective therapeutic agents in treatment of
chemotherapy-induced nausea, vomiting, and irritable bowel syndrome.
Several experimental studies have shown the effect of pharmacological
agents such as 9-tetrahydrocannabinol (THC), the psychoactive
component of Cannabis, Cannabidiol (CBD), a non-psychoactive
ingredient of Cannabis plant, Menthol, Propofol, and etc. on the
functional human 5-HT3 receptors expressed in Xenopus oocyte as well
as rat nodose ganglion neurons [1,2]. 5-HT evoked currents recorded by a
two-electrode clamp technique were inhibited by ligands in a
concentration dependent manner. Simulations of allosteric inhibition
were modeled using Vina docking techniques with the 5-HT3 structure
(see Tables 1 and 2 for results). The 5-HT3 structure was found using
homology sequence similarity techniques with the neuronal nicotinic
acetylcholine receptor (nACH) and inhibitory neurotransmitter receptor
for GABA(A). Results of studies with other members of the superfamily of
ligand gated ion channels signified key residues involved in ligand
binding sites within the transmembrane region of 5-HT3 [3]. Flexible and
rigid docking simulations around key residues resulted in a number of
low-energy (high affinity) configurations of ligand binding (Figure 2). The
predicted residues TYR and THR may constitute a naturally occurring
binding site for 5-HT3.
Conclusion: Experimental inhibition of 5-HT3 shows similar trend in
computational binding energies to the lower binding site (THR 361).
Docking calculations provide explanation of molecular basis of difference
in inhibition by menthol like compounds. Similar binding energies for
THC and CBD corresponds to their similar inhibition of membrane
currents measured in experiment.
References
1. Yang KHS, Isaev D, Morales M, Petroianu G, Galadari S, Oz M: The effect of
9-tetrahydrocannabinol on 5-HT3 receptors depends on the current
density. Neuroscience 2010, 171:40-49.
2. Ashoor A, Nordman JC, Veltri D, Yang KHS, Shuba Y, Kury LA, Bassem
Sadek B, Howarth FC, Shehu A, Kabbani N, Oz M: Menthol inhibits 5-HT3
receptor-mediated currents. J Pharmacol Exp Ther 2013, 347(2):398-409.
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 189 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

Figure 2(abstract P290) Interaction diagrams of Propofol with lower binding site (THR361) of 5-HT3A. A. 2-D Ligand interaction diagram. Residues
are annotated with their 3-letter amino acid code, and their position classification. Hydrophilic interactions: include the hydroxyl group with residue
Asp266 and Met263 (Blue). Hydrophobic interactions: Ile (267, 356, 364, 494), Ser357, Glu360, and Tyr 495 (Red) B. 3-D depiction of interaction

3. Trattnig SM, Harpse K, Thygesen SB, Rahr LM, Ahring PK, Balle T: Discovery
of a novel allosteric modulator of 5-HT3 receptor: Inhibition and
potentiation of Cys-loop receptor signaling through a conserved
transmembrane intersubunit site. Journal of Biological Chemistry 2012,
287(30):25241-25254.
P291
More flexibility for code generation with GeNN v2.1
Thomas Nowotny*, James Turner, Esin Yavuz
CCNR, School of Engineering and Informatics, University of Sussex, Falmer,
Brighton BN1 9QJ, UK
E-mail: t.nowotny@sussex.ac.uk
BMC Neuroscience 2015, 16(Suppl 1):P291
Background: GeNN (GPU enhanced Neuronal Networks) [1,2] is a
software framework that was designed to facilitate the use of GPUs
(Graphics Processing Units) for the simulation of spiking neuronal
networks. It is built on top of the CUDA (Common Unified Device
Architecture) [3] application programming interface provided by NVIDIA
Corporation and is entirely based on code generation: Users provide a
compact description of a spiking neuronal network model and GeNN
generates CUDA and C++ code to simulate it, also taking into account
the specifics of the GPU hardware detected at compile time.
Methods: In this contribution we describe novel work on GeNN, which
has transformed it to a yet more flexible tool for facilitating the use of
GPUs for simulations accelerated by GPUs. The main innovations involve
replacing previous fixed templates for synapse dynamics and learning
models by user-definable code snippets, so allowing redefinition of
virtually every dynamic element of a neural network simulation. This
transition has also enabled the completion of the Brian2 to GeNN and
SpineML to GeNN interfaces [4].
Results: GeNN now allows the free definition of all four, neuron dynamics,
neuron threshold conditions, synapse dynamics and connection weight
dynamics (learning). The desired behavior is encoded in code snippets that
contain C++ compatible code that describes the operations that are
necessary to complete one time step. Table 1 summarizes the available
code slots and their function.
Other improvements in GeNN 2.1 include an improved CUDA block size
estimation algorithm, access to pre- and post-synaptic variables in synaptic
models, and a number of bug fixes.
Conclusion: GeNN has reached level of stability where it should be of
increasing use to the wider computational neuroscience community, in
particular with the completion of its interfaces to other simulators.

Table 1(abstract P291) Summary of code slots available in GeNN for user-defined models
Element Snippet Deployment and Function
Neuron simCode Main time step update of the neuron dynamics
thresholdConditionCode A Boolean expression defining when spikes occur, checked every time step
resetCode The code that defines a change in neuron variables, employed when a spike occurs
Synapse simCode Code that describes the synapse update after a pre-synaptic spike
simCodeEvnt Code that describes the synapse update after a pre-synaptic spike event
simLearnPost Code for the synaptic update triggered by a post-synaptic spike
eventThreshold A Boolean expression that defines synaptic events
synapseDynamics Update code for internal synapse dynamics applied every time step independent of spiking
Post-Synapse postSyntoCurrent Code that describes the transformation of synaptic variables into a post-synaptic current
postSynDecay Code that describes the shared dynamics of summed synaptic activation, typically decay
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Acknowledgements: This work was supported by the EPSRC (Green
Brain Project, grant number EP/J019690/1) and a Royal Academy of
Engineering/Leverhulme Trust Fellowship.
References
1. Nowotny T: Flexible neuronal network simulation framework using code
generation for NVidia CUDA. BMC Neuroscience 2011, 12(Suppl 1):P239.
2. Yavuz E, Turner J, Nowotny T: Simulating spiking neural networks on
massively parallel graphical processing units using a code generation
approach with GeNN. BMC Neuroscience 2014, 15(Suppl 1):O1.
3. CUDA. [[http://www.nvidia.com/object/cuda_home_new.html]], accessed
2015-02-25.
4. Nowotny T, Cope AJ, Yavuz E, Stimberg M, Goodman DFM, Marshall J,
Gurney K: SpineML and Brian 2.0 interfaces for using GPU enhanced
Neuronal Networks (GeNN). BMC Neuroscience 2014, 15(Suppl 1):P148.
P292
Mathematical modelling of ICAN-mediated persistent firing in
hippocampal neurons
Francesco Giovannini1*, Motoharu Yoshida2, Laure Buhry1
1
Neurosys Team, INRIA, LORIA UMR 7503, CNRS, Universit de Lorraine,
Villers-ls-Nancy, F-54600, France; 2Faculty of Psychology, Mercator Research
Group - Structure of Memory, Ruhr-University, Bochum, 44801, Germany
E-mail: francesco.giovannini@inria.fr
BMC Neuroscience 2015, 16(Suppl 1):P292
Persistent neural activity has been the focus of neuroscientific research
since it was first associated with complex cognitive behaviours. In
particular, persistent firing has long been thought to be the neural
mechanisms underlying short-term memory encoding and storage [1]. This
activity is often elicited by short transient stimuli that have to be retained
in memory for long delay periods, in the order of several seconds, after the
original stimulus disappeared. In this scenario, the brain stores information
for future execution of action depending on that information.
Persistent activity elicited in a recurrent network comprising strong
excitation in the local circuit has been extensively described in the
literature (for a review see [2]). However, recent findings have shown that
memory can also be encoded in brain regions which do not display such
recurrent connection topology, including the CA1 hippocampal area [3].
Emerging studies are pointing towards intrinsic neural mechanisms
independent of synaptic connections, as a complementary mechanism for
the maintenance of persistent activity in the hippocampus [4]. These
include various cytoplasmic currents flowing through the membrane,
characterised by slow ion channel kinetics, and particular neurotransmitter
modulation. Our work investigates persistent firing activity in networks of
hippocampal neurons, elicited by leveraging intrinsic currents rather than
network dynamics.
Building on previous findings [5], we model a hippocampal pyramidal
neuron using the Hodgkin-Huxley model, with low-threshold Ca 2+
currents governing the inward flow of calcium ions in the cell membrane.
The intracellular calcium concentration mediates the opening of calcium-
activated non-specific (CAN) ion channels [6], which causes an increase in
the ionic flow through the cell membrane. Therefore the CAN channels
equip the neuron with an after-spike depolarisation mechanism, which
enables it to emit action potentials in the absence of external stimulation.
Given a transient 250ms 200pA current injection, the neuron is capable of
eliciting and maintaining persistent activity with a firing rate of 6Hz for
long delay periods (~30s). Moreover, this behaviour is in accord with that
displayed in neural recordings of hippocampal slice preparations [4].
Connecting these persistent firing neurons in a network comprising
strong local excitation yields a wide range of behaviours depending on
the interaction between CAN and synaptic currents. Indeed, our network
model is capable of displaying rhythmic behaviour in the form of short
synchronised bursts with intra-burst frequencies of 20-40Hz and inter-
burst frequencies of 3Hz. These results hint towards a possible
mechanism for the generation of memory-related oscillatory activity in
the hippocampus.
References
1. Fuster J, Alexander GE: Neuron Activity Related to Short-Term Memory.
Science 1971, 173:652-654.
2. Wang X-J: Synaptic reverberation underlying mnemonic persistent
activity. Trends Neurosci 2001, 24:455-463.
Page 190 of 200
3. Amaral DG: Emerging principles of intrinsic hippocampal organization.
Curr Opin Neurobiol 1993, 3:225-229.
4. Knauer B, Jochems A, Valero-Aracama MJ, Yoshida M: Long-lasting intrinsic
persistent firing in rat CA1 pyramidal cells: a possible mechanism for
active maintenance of memory. Hippocampus 2013, 23:820-831.
5. Destexhe A, Contreras D, Sejnowski TJ, Steriade M: A model of spindle
rhythmicity in the isolated thalamic reticular nucleus. J Neurophysiol 1994,
72:803-818.
6. Partridge LD, Swandulla D: Calcium-activated non-specific cation
channels. Trends Neurosci 1988, 11:69-72.
P293
Spatiotemporal dynamics in spiking simulations of superior colliculus
fit via MCMC suggest disinhibition responsible for superlinear
summation
Richard Veale1*, Tadashi Isa1,2, Masatoshi Yoshida1,2
1
National Institute for Physiological Sciences, Okazaki, Japan; 2SOKENDAI
(The Graduate University for Advanced Studies), Hayama, Japan
E-mail: richard@nips.ac.jp
BMC Neuroscience 2015, 16(Suppl 1):P293
The superior colliculus (SC) is a midbrain region with visually responsive
neurons in the superficial layers and eye movement controlling neurons
in the deeper layers. Recently, [1] performed in vitro experiments to
elucidate lateral interactions within horizontal slices of the SC (Figure 1A).
The experiments indicate that the superficial (visual) layers implement
surround inhibition, and furthermore that strong stimulation at two
adjacent locations (separation ~150 m) produces an unexpected super-
linear summation that is not seen in the deeper layers (Figure 1B, a+b).
We used differential evolution Markov-chain Monte Carlo (MCMC) to
estimate the parameters of a large-scale spiking neural circuit simulation
to fit the slice data [2]. The model contains populations of inhibitory and
excitatory neurons as well as input axons from retina/cortex. We included
free anatomical (dendrite/axon spread) and dynamic (short-term synaptic
plasticity) hyper-parameters in the model, and used MCMC estimate the
posterior distribution of parameters that is most likely given the slice
data. The resulting marginal distributions show promising agreement with
verifiable anatomical parameters such as the lateral spread of dendrites
and axons of the inhibitory and excitatory neuron populations in the
superficial colliculus, even though no such constraints were coded into
the model [2]. However, the posterior distributions for non-intuitive
parameters (such as synaptic efficacies and facilitation/depression time
constants) cannot be verified directly with existing data. Furthermore, it is
not clear what the role of the dynamical parameters is in producing the
behavior of the best-fit models, for example the local superlinearity
described above. In this work, we take the additional step of analyzing
the spatio-temporal dynamics of one of the best-fit regions of parameter
space found via MCMC. The purpose is to provide a mechanistic
explanation for the super-linear summation observed during two-point
stimulation. Figure 1D shows the difference in spatio-temporal dynamics
between actual two-point stimulation versus the linear sum of two
independent stimulations. Thus, positive values indicate an increase in
conductance sent to neurons at the horizontal axis position, arriving from
neurons at the position indicated by the vertical axis (as explained in 1C).
16 ms from stimulation onset, there is a large increase in flow of
inhibitory input from near the center of the circuit to inhibitory neurons
all around the circuit, thus disinhibiting the circuit. 5 ms later, the
excitatory neurons near the center receive input from excitatory neurons
near the middle (i.e. recurrent activity), suggesting that cause for the
super-linearity.
Acknowledgements: RV is funded by JSPS (PE14009). TI funded by MEXT
(26221003), JSPS KAKENHI (13350518). MY funded by a JST/DFG/BMBF
collaborative project in Computational Neuroscience (12800297).
References
1. Phongphanphanee P Marino RA, Kaneda K, Yanagawa Y, Munoz DP, Isa T:
Distinct local circuit properties of the superficial and intermediate layers
of the rodent superior colliculus. Eur J Neurosci 2014, 40:2329-2343.
2. Veale R, Isa T, Yoshida M: Applying Differential Evolution MCMC to
Parameterize Large-scale Spiking Neural Simulations. Proc. IEEE CEC 2015
in press.
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 191 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

Figure 1(abstract P293) (A) Experimental setup of [1], showing slice containing excitatory (red) and inhibitory (gray) neurons. Stimulation applied
to electrodes at lateral distances while recording from an excitatory neuron. (B) Empirical results (dotted) and model fits. Purple: repetitive stimulation
single point, Red: repetitive stimulation two point. (C) Example of a dynamics plot, in this case InhibExcit, showing net conductance in circuit as a
function of sending and receiving neuron. (D) Difference between dynamics plots. Two-point stimulation minus the linear sum of two single-point
stimulation conditions, showing where and when non-linear changes in activity occur

are examined. Sixteen college students were recruited and randomly

P294
Natural environment promotes deeper brain functional connectivity
than built environment
Zheng Chen1*, Yujia He2, Yuguo Yu2*
1
Department of Landscape Studies, College of Architecture and Urban
Planning, Tongji University, Shanghai, 200092; China; 2The State Key
Laboratory of Medical Neurobiology and Institutes of Brain Science, School
of Life Sciences, Fudan University, Shanghai, 200433;China
E-mail: zhengchen@tongji.edu.cn; yuyuguo@fudan.edu.cn
BMC Neuroscience 2015, 16(Suppl 1):P294
Not only genes but also living environments can effectively shape living
human infant brain growth and function performance through learning-
driven neural plasticity. However, few evidences demonstrated that
exposure to different environments may modulate adult brain cognitive
functions [1]. This study examined this issue by using the Emotiv EPOC
wireless EEG headset [2] and accompanying software. Brain waves are
measured in terms of amplitude (10-100 microvolts) and five frequency
bands, i.e., (0.5-4 Hz), (4-8 Hz), a (8-12 Hz), b (13-25 Hz) and g (25-70),
assigned to the two conditions. Participants were asked to sit either in a
built environment (i.e., a traffic island under an elevated highway), or in a
natural environment (i.e., a heavily wooded campus garden). They were
first sitting facing walls as baselines excluded for visual exposure for eight
one-minute sessions with their eye open and closed in turns (OCOCOCOC),
and then they turned to scene facing and exposed to the environment for
20 minutes. EEG was measured in the latter 10 minutes of exposure, as
well as during eye-open and eye-closed baseline sessions.
Functional connectivity analysis revealed that subjects with eye close in
both environments have stronger or deeper functional connectivity
among different brain regions than eye-opened cases, while eye-opened
subjects walking in natural environments have stronger functional
connectivity than in highway environments (see Figure 1a). Interestingly,
power spectrum analysis showed that EEG powers in all the frequency
bands are higher in natural environment than in built environment (see
Figure 1b~1f), indicating large-amplitude synchronized EEG waves in the
brains of natural environment which strengthen deeper functional
connectivities among brain regions (see Figure 1g,h for example). These

Figure 1(abstract P294) Results of the exposure to built environment (B) and that to natural environment (N), were reported in mean power of
band (b), band (c), a band (d), b band (e) and g band (f). Average Cross-correlations of the 12 channels were displayed in (a). Cross-correlations of
mean b powers between channels for to built environment (g) and natural environment (h) ,respectively. Wall-facing eye-closed pretest baselines were
marked as -C, while post-exposure statuses were marked as -P
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
results suggest that natural environment may promote a better brain
performance than built environment.
Acknowledgements: This project is funded by the National Natural
Science Foundation of China (31271170, 51408429), Shanghai Eastern
Scholar SHH1140004 program and Pujiang Program 14PJC099
References
1. Aspinall P, Coyne R, Roe J: The urban brain: analysing outdoor physical
activity with mobile EEG. Br J Sports Med 2015, 49:272-276.
2. Dekihara H, Iwaki T: Development of human computer interface based
on eye movement using Emotiv EEG headset. International Journal of
Psychophysiology 2014, 94(2):188-188.
P295
Axon initial segment potassium channel density in cortical neurons
Wen Zhang, Boqiang Fan, Ping Zheng, Yuguo Yu*
The State Key Laboratory of Medical Neurobiology and Institutes of Brain
Science, School of Life Sciences, Fudan University, Shanghai, 200433;China
E-mail: yuyuguo@fudan.edu.cn
BMC Neuroscience 2015, 16(Suppl 1):P295
There is a growing interest in estimating actual density ranges of Na+
channels in the very thin axon, especially in the action potential (AP)
initiation zone, i.e., the axon initial segment (AIS, 20-50 microns away
from the cell body). Both immunostaining studies and patch-clamp
recordings indicated a relatively high density of Na+ channels in AIS of
either pyramidal regular-spiking (RS) cells [1] or fast-spiking (FS)
GABAergic interneurons [2,3]. Here, we investigated potassium channel
densities in AISs of both RS and FS cells in same recording conditions.
Our axonal recordings directly revealed that there is a very lower
potassium density gK = 185.819 pS/m2 N = 16) for the RS AIS while a
Figure 1(abstract P295) A. The bar graph shows peak K+ conductance density recorded in axon initial segment of cells (16 recording axons for
RS pyramidal cell, 11 for FS interneuron) by outside-out axon patch recording technique. B. Summary of the results from partially blocking K+
channels. The AP dV/dt ratio is an exponentially decaying function of the spike duration
Page 192 of 200
higher gK (495.7108 pS/m2, N = 11) for FS AIS, see Figure 1A. For both
the RS pyramidal cells and FS PV cells, partially blocking K+ channels by
applying 4-AP broadened the spike duration and decreased the dV/dt
ratio significantly (P < 0.05) (For RS cells: N = 5; For RS cells: N = 4).
Interestingly, we observed that the AP dV/dt ratio is an exponentially
decaying function of the spike duration for both RS- and FS-spikings (see
Figure 1B), such that y = 0.12+0.16EXP((0.5-x)/0.2), where y represents the
dV/dt ratio and represents the AP duration. These observations suggest
strongly that potassium channel density is one of the major intrinsic
factors dominating the spike shape properties, especially half-height spike
duration and dV/dt ratio.
In sum, the significant difference in potassium channel density in axonal
initial segment where action potentials are initiated may play a critical
role in controlling action potential properties of both RS- and FS-spiking
cells in nervous system by the same general biophysical rule. These
results may be important for constructing computational models of
different types of cortical neurons.
Acknowledgements: This project is funded by NNSF of China (31271170)
and Eastern Scholar SHH1140004) at Shanghai Institutions of Higher
Learning
References
1. Hu W, Tian C, Li T, Yang M, Hou H, Shu Y: Distinct contributions of Na(v)
1.6 and Na(v)1.2 in action potential initiation and backpropagation. Nat
Neurosci 2009, 12(8):996-1002.
2. Hu H, Jonas P: A supercritical density of Na(+) channels ensures fast
signaling in GABAergic interneuron axons. Nat Neurosci 2014,
17(5):686-693.
3. Li T, Tian C, Scalmani P, Frassoni C, Mantegazza M, Wang Y, Yang M, Wu S,
Shu Y: Action potential initiation in neocortical inhibitory interneurons.
PLoS Biol 2014, 12(9):e1001944.
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 193 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

P296
Key factors dominating the neural coding preference to 1/f signal
Boqiang Fan, Wen Zhang, Shanglin Zhou, Yuguo Yu*
School of Life Sciences, the State Key Laboratory of Medical Neurobiology
and Institutes of Brain Science, Fudan University, Shanghai, 200433, Republic
of China
E-mail: yuyuguo@fudan.edu.cn
BMC Neuroscience 2015, 16(Suppl 1):P296
Experiments have demonstrated that cortical and sensory neurons prefer
to response to signals with characteristics of long-term correlation or 1/f
noise feature better than signals with no correlation like white-noise-type
[1]. In order to study the underlying mechanism, we built up a cortical
neuronal model [2] based on Hodgkin-Huxley theory to study the
correlations between neuron kinetics and signal statistics. Interestingly, we
observed that (see Figure 1a,b) white-noise-type signal (cutoff frequency
>10000Hz) is hard to induce action potentials unless signals with very
strong intensity while 1/f signal and low-pass filtered white noise type
signal (cutoff frequency <1000Hz) can easily induce action potentials with
high firing rate at low signal intensity (quantified by signal standard
deviation STD), see Figure 1a. Moreover, the half-height duration of action
potentials is also varying with more sensitivity to 1/f signal or filtered white
noise than pure white noise (Figure 1b). In addition, neural firing rate and
spike duration are more sensitive to 1/f signal than filtered white noise.
This is a clear positive evidence of neuronal coding preference to signals
with long-term correlations.
To further reveal the key factors dominating the preference of neuronal
dynamics to colored and white noise, we systematically varied the values
of Na+ and K+ channel time constants and channel rate constants. Figure
1c showed that the firing rate doesnt change much as a function of
sodium activation time constant (m) for a given 1/f signal (STD = 1).
However, the firing rate decrease dramatically with an increase of sodium
inactivation time constant (h) or potassium activation time constant (n)
for this signal. For very large h and n, neuronal response dynamics start
to be saturate for all signals, see Figure 1d. These results indicate that
both sodium inactivation time constant (h) and potassium activation
time constant (n) may be the key factors dominating neural coding
preference to signals with different correlated statistic features.
In sum, this study demonstrated that ionic channel time constant of
sodium inactivation and potassium activation may be the dominating
factors accounting for the neuronal spiking sensitivity favorable to signals
with temporal correlations.
Acknowledgements: This project is funded by NNSF of China (31271170)
and Shanghai Eastern Scholar program SHH1140004)
References
1. Yu Y, Romero R, Lee TS: Preference of sensory neural coding for 1/f
signals. Phys Rev Lett 2005, 94:108103.

Figure 1(abstract P296) Interactions among neuron kinetics and signal statistics. a) Firing rate vs signal STD for white noise, filtered white noise and
1/f noise. b) Half-height duration of action potentials vs signal STD for white noise, filtered white noise and 1/f noise. c) For a fixed input 1/f signal (STD =
1), firing rate change as a function of sodium activation time constant (m), inactivation time constant (h) and potassium activation time constant (n). d)
firing rate vs signal STD for , In the situation with both h and n being increased to 15 times larger
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
2. Yu Y, Hill AP, McCormick DA: Warm body temperature facilitates energy
efficient cortical action potentials. PLoS Comput Biol 2012, 8(4):1-16.
P297
Learning experience facilitates sparse coding of new odors in a large-
scale olfactory bulb model
Shanglin Zhou1*, Boqiang Fan1, Michele Migliore2, Yuguo Yu1*
1
School of Life Sciences, the State Key Laboratory of Medical Neurobiology
and Institutes of Brain Science, Fudan University, Shanghai, 200433, China;
2
Institute of Biophysics, National Research Council, 90146 Palermo, Italy
E-mail: yuyuguo@fudan.edu.cn
BMC Neuroscience 2015, 16(Suppl 1):P297
Odor responses of mitral cells in the olfactory bulb are observed to be
sparse spatially and decorrelated in response to odor signals [1]. We have
built a large scale biophysical network model of olfactory bulb composed
of mitral and granule cells, corresponding to 1/100th of the real system
in the rat, and used direct experimental imaging data of glomeruli
activated by various odors. Our previous reports have showed that a
sparse spatial spiking representation of specific odor signals can emerge
naturally within several seconds learning period from the mitral-granule
cell interactions, realistically implemented in our model with self-
organizing dendrodendritic synapses driven by mitral cell activity [2]. To
address how the prior odor experience interfere with subsequent sparse
coding of new coming odors, we trained the network with a set of odors
(Figure 1a and 1b, note that no signal in left panel of Figure 1b which is
called native state) to reach a stable response state (sniff frequency
around 4-10 Hz) within 5 seconds (Figure 1a), and then a second odor (e.
g., octanol 8OH) was input to the network (5s, 4-10 Hz) (Figure 1b,
middle and right). Compared to the network trained from the native state
(Figure 1b, left), prior active odor training could significantly facilitate the
whole mitral cell network into different degrees of response sparseness
for odor 8OH (see right two panels in Figure 1b). The sparseness onset
time S1/2 (defined as the time reached the half of maximum sparseness)
is relatively large for input signals with less similarity (Figure 1c) while
become shorter as more input similarity is induced (Figure 1d, correlation
Figure 1(abstract P297) (a) odor inputs 8OH, 5CHO and k7-1. (b) MC network response to odor 8OH for nave training (left), 5CHO prior training
(middle) and k7-1 training (right). (c) Network sparseness develops for different sets of prior odor training. (d) Sparseness onset time S1/2 vs. input signal
correlation coefficient (corrcoef.). (e) Mean corrcoef between MCs within the network vs. input signal corrcoef. (f) Response similarity (quantified by
corrcoef between MC network responses to different odor inputs) vs input signal corrcoef
Page 194 of 200
coefficient r = 0.89), however, all shorter than that in native situation
(Figure 1c, d), suggesting prior training experience could significantly
promote odor perception of even novel odors.
Furthermore, for the most of total 16 prior odor training tests, the mean
correlation coefficients between MT cells within the network measured at
the end of after-learning process (i.e. No.14-15s) are less than the native
situation (Figure 1e), suggesting a more decorrelated state. Correspondingly,
the similarity of the MT network responses for most of the tests is lower
significantly than the nave case (Figure 1f). This strongly suggests prior odor
experience could significantly promote odor discrimination ability of the
olfactory bulb network. In sum, odor experience could accelerate the
formation of network response sparseness, decorrelation, and promote
signal discrimination.
Acknowledgements: This project is funded by NNSF of China (31271170)
and Shanghai Eastern Scholar program SHH1140004).
References
1. Kato Hiroyuki K, Chu Monica W, Isaacson Jeffry S, Komiyama T: Dynamic
Sensory Representations in the Olfactory Bulb: Modulation by
Wakefulness and Experience. Neuron 2012, 76(5):962-975.
2. Yu Y, Migliore M, Hines ML, Shepherd GM: Sparse coding and lateral
inhibition arising from balanced and unbalanced dendrodendritic
excitation and inhibition. J Neurosci 2014, 34(41):13701-13713.
P298
Axon zippering in neuronal cell culture and its biophysical modeling
Daniel Smit1,2,3, Coralie Fouquet3, Frederic Pincet4, Alain Trembleau3,
Martin Zapotocky1,2*
1
Institute of Physiology of the Czech Academy of Sciences, Prague, Czech
Republic; 2Institute of Biophysics and Informatics, First Faculty of Medicine,
Charles University in Prague, Prague, Czech Republic; 3IBPS, Neuroscience
Paris Seine, CNRS UMR8246, Inserm U1130, UPMC UM 119, Universit Pierre
et Marie Curie, Paris, France; 4Laboratoire de Physique Statistique, Ecole
Normale Superieure, Paris, France
E-mail: zapotocky@biomed.cas.cz
BMC Neuroscience 2015, 16(Suppl 1):P298
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
During neural development, the growing axons frequently adhere to each
other and form tight bundles. This process contributes to the setting up
of correct connectivity in, e.g., the mammalian olfactory system [1]. Here
we report an experimental and theoretical investigation of axon bundling
in olfactory epithelium explants; in this system, the dynamics of axon-
axon interactions can be directly examined.
Our analysis is based on time-lapse optical microscopy observations of
axon growth in primary cell culture. Olfactory epithelium explants from
mouse embryos (day 13-14) were cultured on laminin substrate for two
days and then recorded using DIC videomicroscopy for up to 24 hours.
The growing axons established a dense network within which large
bundles of axons were progressively formed. This dynamics was driven
by numerous fast zippering processes, during which the length of the
segment along which two axons adhere to each other was increased.
Similar processes were inferred in 1982 from electron-microscopy-based
analysis of sensory neurites in Xenopus embryos [2], but have been very
rarely studied in the subsequent literature.
We digitized a collection of 30 spontaneous zippering processes observed
in the developing axon network. During zippering, the vertex (i.e. the
point where the two axons meet) advances with a typical velocity of 0.2
to 1 m/min. The distance between immediate position of the vertex and
the final equilibrium position decreases either linearly with sudden arrest
or exponentially in time. We observed similar behaviour, at velocities of 1
to 3 m/min, for zippering events induced during micro-manipulation
experiments.
To explain these observations, we developed a simplified biophysical model
that includes the adhesive force between the axons and the mechanical
tension in the axons, as well as adhesion and anisotropic friction with the
substrate. A static zipper is obtained if the tensile and adhesive forces are
balanced. The model predicts that the linear regime of convergence to
equilibrium can arise from a perpendicular outer force acting on one of the
axons, while the exponential regime is a result of sudden change in axial
tension of the axon. The parameters of the model were estimated from
calibrated micro-manipulations of the culture, using a biotinylated red blood
cell as force transducer [3]. These experiments allowed us to measure the
tension within the axon shaft (typically about 1 nN) and deduce the force of
adhesion between two axon shafts (of order 100 pN).
To account for additional factors, we implemented a molecular-dynamics-
type simulation. The beads-and-springs representation of each axon permits
to include bending rigidity and allows for an inhomogeneous distribution of
tension along the axon. To implement the interaction between axons, we
adapted results from contact mechanics [4] for cylindrical bodies. This
permits to avoid artifacts that would result from direct adhesive interaction
between beads. We compare the dynamics of zippering in this detailed
model to the predictions of the simplified analytical model.
Acknowledgements: Work supported by GAR 14-16755S, GAUK
396213, MMT 7AMB12FR002, SVV-260023/2014, NIH 1RO1DCO12441 and
ANR 2010-BLAN-1401-01.
References
1. Miller AM, Maurer LR, Zou DJ, Firestein S, Greer CA: Axon fasciculation in
the developing olfactory nerve. Neural Dev 2010, 5:20.
2. Roberts A, Taylor JSH: A scanning electron microscope study of the
development of a peripheral sensory neurite network. J Embryol Exp
Morph 1982, 69:237-250.
3. Pincet F, Husson J: The solution to the streptavidin-biotin paradox: the
influence of history on the strength of single molecular bonds. Biophys J
2005, 89:4374-4381.
4. Johnson KL, Kendall K, Roberts AD: Surface energy and the contact of
elastic solids. Proc Roy Soc London A 1971, 324:301-313.
P299
Computational estimation of calcium fluxes in isolated magnocellular
neurons
S Kortus1,2,3, G Dayanithi3,4, M Zapotocky1,2*
1
Institute of Physiology of the Czech Academy of Sciences, Prague, Czech
Republic; 2Institute of Biophysics and Informatics, First Faculty of Medicine,
Charles University in Prague, Prague, Czech Republic; 3Institute of
Experimental Medicine, Czech Academy of Sciences, Prague, Czech Republic;
4
INSERM U710/EPHE, Universit Montpellier 2, Montpellier, France
E-mail: zapotocky@biomed.cas.cz
BMC Neuroscience 2015, 16(Suppl 1):P299
Page 195 of 200
Current optical methods based on fluorescent indicators permit to
measure the intracellular calcium concentration with a high temporal
resolution. To analyze the physiological mechanisms that underly the
calcium dynamics, however, knowledge of the calcium fluxes into and
out of the cell is needed. Here we present a method that permits to
separately estimate the influx and clearance rates, based on the
measurement of Ca 2+ concentration during a series of depolarization-
evoked calcium transients. We apply this method to investigate calcium
clearance mechanisms in isolated magnocellular neurons of the rat
supraoptic nucleus.
In the simplest case, we assume that the cytoplasmic Ca2+ concentration
during the transient is governed by a time-dependent influx J influx (t)
through voltage-gated Calcium channels and by an outflux Jclearance that
depends only on the instantaneous calcium concentration [Ca2+](t):
d[Ca2+ ]/dt = Jinflux (t) Jclearance ([Ca2+ ]. (1)
To separate the two fluxes, we first estimate the clearance function
Jclearance([Ca2+]). Near the end of the transient, Jclearance dominates over
Jinflux, and Jclearance may be obtained directly as the measured Ca2+ decay
rate [1]. In contrast, near the peak of a transient the two fluxes are
comparable, and J clearance therefore significantly exceeds -d[Ca 2+ ]/dt.
However, in this case the clearance rate obtained from a higher transient
can be used as a good estimate. If the assumption of Eq.1 is satisfied, the
clearance function is obtained as the envelope of the recorded return
curves in the d[Ca2+]/dt vs. [Ca2+] plot. The calcium influx rate Jinflux(t)
during each transient is then estimated by substracting Jclearance([Ca2+](t))
from the measured rate d[Ca 2+ ]/dt. We tested the adequacy of this
procedure using surrogate calcium dynamics data.
For cells in which the endoplasmic reticulum (ER) noticeably contributes
to the calcium transient [2], the clearance function Jclearance is not solely
dependent on the cytoplasmic calcium concentration, as was assumed
above. In this case, the method described above can still be applied to
experiments performed in presence of Thapsigargin or cyclopiazonic acid
(CPA), to avoid release or uptake of Ca2+ by the ER. Comparison of the
estimated fluxes from the experiments with and without Thapsigargin/
CPA can be used to investigate the ER-dependent calcium fluxes. We apply
this method to freshly isolated magnocellular neurons, in which we used
Fura-2AM to measure the cytoplasmic [Ca2+] during depolarization-evoked
Ca2+ transients of various amplitudes and durations; depolarization was
induced [3] by changing the external K+ concentration.
Our method of estimating the Ca2+ fluxes may be used also in other cell
types to help characterize the contribution of individual mechanisms to
calcium dynamics.
References
1. Fierro L, DiPolo R, Llano I: Intracellular calcium clearance in Purkinje cell
somata from rat cerebellar slices. Journal of Physiology 1998,
510(2):499-512.
2. Gilon P, Arredouani A, Gaill P, Gromada J, Henquin JC: Uptake and release
of Ca2+ by the endoplasmic reticulum contribute to the oscillations of
the cytosolic Ca2+ concentration triggered by Ca2+ influx in the
electrically excitable pancreatic B-cell. J.Biol.Chem 1999, 274:20197-20205.
3. Harasztosi C, Por A, Rusznak Z, Szucs G: Removal of Ca2+ following
depolarization-evoked cytoplasmic Ca2+ transients in freshly dissociated
pyramidal neurones of the rat dorsal cochlear nucleus. Brain Research
2002, 930:123-133.
P300
Functional identification of complex cells from spike times and the
decoding of visual stimuli
Aurel A Lazar*, Nikul H Ukani, Yiyin Zhou
Department of Electrical Engineering, Columbia University, New York, NY
10027, USA
E-mail: aurel@ee.columbia.edu
BMC Neuroscience 2015, 16(Suppl 1):P300
Neural circuits built with complex cells play a key role in modeling the
primary visual cortex. The encoding capability of an ensemble of complex
cells has not been systematically studied, however. Can visual scenes be
reconstructed using the spike times generated by an ensemble of
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 196 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

Figure 1(abstract P300) A. SNR of reconstruction when varying the number of measurements/spikes (dimension of the input space: 117).
B. Mean SNR of identified complex cell DSP when varying the number of input trials used in identification. C. Comparison of identification by Complex
Cell CIM with STC. Quadrature pair Gabor filters (1st column) identified with Complex Cell CIM with 746 spikes (2nd column, SNR: 123.08 [dB], 88.93 [dB]),
and with STC using 39, 769 spikes (3rd column, SNR: 16.77 [dB], 17.90 [dB]), and using 746 spikes (4th column, SNR: 0.21 [dB], 0.52 [dB]). D. Evaluating
identification quality in the input space. SNR of reconstruction of novel stimuli assumed to be encoded with the identified DSPs

complex cells? Can the processing taking place in complex cells be
identified with high accuracy? Processing by complex cells has the
complexity of Volterra models [1]. General Volterra based models call,
among others, for efficient functional identification and decoding
algorithms.
We demonstrate that complex cells exhibit Volterra dendritic stimulus
processors (Volterra DSPs) that are analytically and computationally
tractable. Decoding and identification problems arising in neural circuits
built with complex cells can be efficiently solved as rank minimization
problems [2]. We provide (i) an algorithm that reconstructs the visual
stimuli based on the spike times generated by circuits with widely
employed complex cells models (Complex Cell Time Decoding Machines),
and (ii) propose a mechanistic algorithm for functionally identifying the
processing in complex cells using the spike times they generate
(Complex Cell Channel Identification Machines). These algorithms are
based on the key observation that the functional identification of
processing in a single complex cell is dual to the problem of decoding
stimuli encoded by an ensemble of complex cells.
In addition, we show that the number of spikes needed for perfect
reconstruction of a band-limited stimulus is proportional to the
dimension of the stimulus space rather than the square of its dimension,
thereby reducing the required number of neurons/measurements to a
physiologically plausible range. This result demonstrates that visual
stimuli can be efficiently reconstructed from the amplitude information
carried in the complex cells alone. Similar results obtained for
identification establish the computational tractability of higher order
Volterra DSPs. We provide examples of perfect reconstruction of space-
time stimuli (Figure 1A) and examples of identification of complex cell
DSPs (Figure 1B). We demonstrate that our identification algorithms
substantially outperform algorithms based on spike-triggered covariance
(STC) (Figure 1C). Finally, we evaluate our identification algorithms by
reconstructing novel stimuli in the input space using identified Volterra
DSPs (Figure 1D) [3].
Acknowledgements: The research reported here was supported by
AFOSR under grant #FA9550-12-1-0232.
References
1. Lazar AA, Slutskiy YB: Spiking neural circuits with dendritic stimulus
processors: encoding, decoding, and identification in reproducing kernel
Hilbert space. J Comput Neurosci 2015, 38(1):1-24.
2. Cands EJ, Strohmer T, Voroninski V: PhaseLift: Exact and stable signal
recovery from magnitude measurements via convex programming.
Comm Pure Appl Math 2013, 66(8):1241-1274.
3. Lazar AA, Slutskiy YB, Zhou Y: Massively parallel neural circuits for
stereoscopic color vision: Encoding, decoding and identification. Neural
Networks 2015, 63:254-271.
P301
Retina of the fruit fly eyes: a detailed simulation model
Aurel A Lazar*, Konstantinos Psychas, Nikul H Ukani, Yiyin Zhou
Department of Electrical Engineering, Columbia University, New York, NY
10027, USA
E-mail: aure@lee.columbia.edu
BMC Neuroscience 2015, 16(Suppl 1):P301
The compound eyes of the fruit flies consist of the retina and 4 neuropils.
The retina comprises about 700~800 ommatidia, each of which houses 8
photoreceptors, where the phototransduction takes place. Although
seemingly simple, each photoreceptor can perform complex computation
on their own [1]. Nevertheless, it is not yet clear how individual
photoreceptors contribute to an overall spatiotemporal processing of
visual scenes. Towards this goal, we constructed a full-scale simulation of
the retina on a cluster of GPUs.
The geometry of the eye was taken into account by building a hexagonal
array of 721 ommatidia on a hemisphere and assigning a proper optic axis
and an acceptance angle to each of the photoreceptors. The visual
stimulus was presented on either a hemisphere or a cylinder surrounding
the eye, as in many experimental settings (see Figure 1).
The phototransduction process is based on a biochemical model of 30,000
microvilli followed by a conductance based biophysical model of the cell
membrane [1]. Here we only consider monochromatic photoreceptors R1-R6
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Figure 1(abstract P301) A. Stimulus presented on a hemisphere screen. B. Inputs (Number of Photons) to R1 photoreceptors in all ommatidia. C.
Geometric relation between eye and screen. D. Gamma corrected screen intensity. E. Gamma corrected inputs to R1 photoreceptors. F. Voltage response
of R1 photoreceptors. G. Averaged voltage response of photoreceptor terminals in all cartridges. H. Response of L1 neurons in all cartridges. I. Response of
L2 neurons in all cartridges
in each ommatidium. Each photoreceptor is described by ~ 450, 000
equations; the total number of equations simulated for the entire retina
amounted to about 1.95 billion. All the simulations were performed on 4
Tesla K20X GPUs; it took approx. 7 minutes to finish 1 second of simulation.
After constructing the detailed simulation model, we tested the
processing of visual stimuli by the retina circuit. Furthermore, to test the
effect of feedback originating in the lamina on the photoreceptors, we
linked the retina model with a lamina model [2] in the Neurokernel
environment [3]. An example of response of the circuit to natural stimuli
is shown in Figure 1. The simulations reveal that composition rules in the
lamina circuit affect the spatiotemporal processing carried out by the
photoreceptors under the model parameters considered here.
Finally, we scaled up the retina model of the fruit fly to the size of the
blow fly to investigate the effect of visual acuity and noise.
Acknowledgements: The research reported here was supported by
AFOSR under grant #FA9550-12-1-0232.
References
1. Song Z, Postma M, Billings SA, Coca D, Hardie RC, Juusola M: Stochastic,
adaptive sampling of information by microvilli in fly photoreceptors.
Curr Biol 2012, 22:1371-1380.
2. Lazar AA, Ukani NH, Zhou Y: The cartridge: A canonical neural circuit
abstraction of the lamina neuropil - construction and composition rules.
Neurokernel RFC #2 2014.
3. Lazar AA, Givon LE: Neurokernel: An open scalable architecture for
emulation and validation of Drosophila brain models on multiple GPUs.
Neurokernel RFC #1 2014.
Page 197 of 200
P302
Network heterogeneity and seizure generation
Sima Mofakham1*, Christian G Fink2, Victoria Booth3, Michal R Zochowski1,4,5
1
Biophysics program, University of Michigan, Ann Arbor, MI , USA;
2
Department of Physics & Astronomy and Neuroscience Program, Ohio
Wesleyan University, Delaware, OH, USA; 3Departments of Mathematics and
Anesthesiology, University of Michigan, Ann Arbor, MI, USA; 4Department of
Physics, University of Michigan, Ann Arbor, MI, USA; 5The R.B. Zajonc Institute
for Social Studies, Stawki , Warsaw, Poland
E-mail: mofakham@umich.edu
BMC Neuroscience 2015, 16(Suppl 1):P302
It has been shown that seizures occur more frequently at the transition
from wake to sleep, or from one stage of sleep to another. Acetylcholine
(ACh) is a neuromodulator that controls wake and sleep stages, and is
present at high levels during waking and is absent in slow wave sleep
(SWS). ACh has also been shown to switch the excitability as measured
with Phase response curves (PRC) of pyramidal cells from Type 2 to Type 1.
In general, Type 1 neurons are integrating type with high-excitability while
Type 2 have lower excitability but higher capacity for synchronization. We
investigate the effect of non-uniform cholinergic modulation, such as
might occur at sleep/wake transitions, on the propensity for neuronal
synchronization in large-scale networks of Hodgkin-Huxley models for
cortical pyramidal cells. The interplay between the cellular properties and
network connectivity in a heterogeneous network of Type 1 and Type 2
neurons can strongly affect network spatio-temporal dynamics. The focus
of this research is to detect conditions that promote synchrony and seizure
BMC Neuroscience 2015, Volume 16 Suppl 1 Page 198 of 200
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1

like activity in a mixed network of Type 1 and Type 2 neurons. Here we
investigate inhomogeneous networks built of neurons that have non-
identical connectivity properties. Namely every cell has an individual ratio
of local and long distance synaptic connections. We show that even if the
structure of the network is identical (i.e. identical adjacency matrix) there is
a differential network-wide synchronization propensity depending on
which neurons have Type II cellular properties.
P303
The role of adaptation current in synchronously firing inhibitory neural
networks with various topologies
Scott Rich1*, Victoria Booth1, Michal Zochowski2
1 consider networks of three cell types: Type I neurons and Type II neurons
Department of Mathematics, University of Michigan, Ann Arbor, MI 48104,
USA; 2Department of Applied Physics, University of Michigan, Ann Arbor,
MI 48104, USA
E-mail: sbrich@umich.edu
BMC Neuroscience 2015, 16(Suppl 1):P303
Inhibitory neural networks have the capacity to fire synchronously
depending strongly on synaptic current dynamics. Various types of
inhibitory interneurons, including some with adaptation currents, are
present in hippocampal circuits and are implicated in governing overall
network pattern formation. However, the contribution of intrinsic cell
firing patterns of interneurons to these patterns has not been fully
investigated. Understanding how both synaptic and cellular properties
contribute to the propensity for inhibitory neural networks to synchronize
is thus an invaluable tool for investigating hippocampal network pattern
formation. Through numerical simulation of large, spiking neuron,
inhibitory networks, we investigate the role of the slow, adapting M-type
K+ current in network pattern formation. This adapting current plays a
role in hippocampal interneurons, including the OLM cells, in which the
blockade of M-current by acetylcholine or other neurotransmitters
switches the neuronal firing rate-current relation (f-I curve) from Type II
to Type I. Other types of interneurons, such as fast-spiking PV cells,
display Type II f-I curves without any adaptation current. Thus, we
that either contain [1] or do not contain an M-type K+ adaptation
current. All cell types are modeled in the Hodgkin-Huxley formalism.
Heterogeneity is introduced to the networks through randomized
external applied current to the neurons. We vary network connection
topologies using the Small World Network Paradigm [2] in order to
systematically investigate the role of connectivity between local and
random topologies.
To probe the interaction of cellular and synaptic properties influencing
synchronization in these networks, we vary the time constant of decay of

Figure 1(abstract P303) Differential synchronization properties (scored by the Golomb measure [3]) of two inhibitory networks as a function of
synaptic current duration, external drive and synaptic strength. In 1000 neuron networks with 300 random incoming synapses per neuron, Type II
networks without an adaptation current show clustering, resulting in a moderate synchronization measure, for nearly all parameters with low synaptic
weight (A). Higher synaptic weight diminishes clustering (C). In identical simulations with low synaptic weight in Type II networks with an adaptation
current, full synchronization is displayed when the synaptic decay constant is in an optimal range that grows with the external input current (B). Stronger
synaptic weight reduces the synchronization parameter range towards shorter lasting synaptic currents (D)
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
the inhibitory synaptic current and the intrinsic cellular frequencies by
varying the mean of the distribution of external currents applied to each
neuron. With sparse, nearest neighbor connection topologies, Type I
networks exhibit stationary activity patterns reminiscent of standing
waves while Type II neurons exhibit traveling wave activity patterns that
sweep across the network. These traveling waves are much more robust
and regular when the Type II neurons contain the adaptation current.
With sparse random connectivity, Type II networks without an adaptation
current exhibit cluster-firing patterns, in which cells segregate into
multiple clusters that show synchrony within the cluster but not across
clusters. In contrast, Type II networks with the adaptation current display
full synchronization for some parameter ranges, but do not exhibit robust
clustering (Figure 1). Preliminary analysis suggests that some of these
results are due to the differences between Type I and Type II f-I profiles
and that the adaptation current may enhance the effects of these
differences.
References
1. Fink C, Booth V, Zochowski M: Cellularly-driven differences in network
synchronization propensity are differentially modulated by firing
frequency. PLoS Computational Biology 2011, 7.
2. Watts D, Strogatz S: Collective dynamics of small-world networks. Nature
1998, 393:440-442.
3. Golomb D, Rinzel J: Dynamics of globally coupled inhibitory neurons
with heterogeneity. Phys. Rev. E 1993, 48:4810-4814.
P304
Modeling the formation and dynamics of cortical waves induced by
cholinergic modulation
James P Roach1*, Eshel Ben-Jacob2,3, Leonard M Sander4,
Michal R Zochowski1,4,5
1
Neuroscience Graduate Program, University of Michigan, Ann Arbor, MI,
48109, U.S.A; 2School of Physics and Astronomy, Tel-Aviv University, Tel Aviv,
69978, Israel; 3Center for Theoretical Biological Physics, and Department of
Biochemistry and Cell Biology, Rice University, Houston, TX, 77005, USA;
4
Department of Physics & Center for Studies of Complex Systems, University
of Michigan, Ann Arbor, MI, 48109, USA; 5Biophysics Program, University of
Michigan, Ann Arbor, MI, 48109, USA
E-mail: roachjp@umich.edu
BMC Neuroscience 2015, 16(Suppl 1):P304
States of arousal, or consciousness with the brain are regulated largely by
the neurotransmitter acetylcholine (ACh). Specifically, ACh is likely
responsible for the transition between slow wave sleep (SWS; where ACh
is absent) and rapid eye movement sleep or waking states (where ACh is
high). Patterns of neural activity within the cerebral cortex corresponding
to these states are markedly different. During SWS there are traveling
waves of intense activity in the cortex while in other states locally
organized stationary patterns occur [1]. From a functional perspective,
stationary patterns are likely to be important for working memory and
attention dynamics while traveling waves could lead to synaptic
renormalization [2]. The mechanism for how changes on the cellular level
are translated to patterns on the network level is not understood. In this
work we give a model for the action of ACh on a network of neurons of
the Hodgkin-Huxley type with a current that is regulated by ACh that
induces spike-frequency adaptation (SFA) [3]. The cells are coupled in a
center-surround scheme. When SFA is minimal (such as in waking or REM
sleep state, high ACh) patterns of activity are localized and easily pinned
to regions defined by enhanced recurrent excitation. Increasing the level
SFA is present (by increasing ACh), traveling waves of activity naturally
arise. Depending on the strength of inhibitory coupling within the
network, SFA is able to induce a wide variety of dynamical regimes
(Figure 1). We present a detailed mechanism that shows that the level of
inhibition sets the spatial extent of network activity and that SFA defines
the temporal scope, which is directly modulated by ACh in the model.
These model calculations give unique insights into the role and
significance of ACh in determining patterns of cortical activity and
functional differences arising from these patterns.
Acknowledgements: This work is supported by the NSF Graduate
Research Fellowship Program under Grant No. DGE 1256260 (JPR), the
Tauber Family Funds and the Maguy-Glass Chair in Physics of Complex
Systems at Tel Aviv University (EBJ), NSF Center for Theoretical Biological
Page 199 of 200
Figure 1(abstract P304) An illustration of the dynamics sampled by
scannig inhibitory strength,(wi e), and gKs. In this model gKs is
increased to simulate decreasing ACh levels. In a general sense, the
spatial scope of activity is determined by the excitatory/ inhibitory
balance, and the temporal scope of activity is determined by the
strength of SFA
Physics Grants PHY-1427654 and NSF-MCB-1214457 (EBJ), NSF CMMI
1029388 (MRZ), and NSF PoLS 1058034 (MRZ \& LMS).
References
1. Massimini M, Huber R, Ferrarelli F, Hill S, Tononi G: The Sleep Slow
Oscillation as a Traveling Wave. Journal of Neuroscience 2004,
24:6862-6870.
2. Tononi G, Cirelli C: Sleep and synaptic homeostasis: a hypothesis. Brain
Research Bulletin 2003, 62:143-150.
3. Stiefel KM, Gutkin BS, Sejnowski TJ: The effects of cholinergic
neuromodulation on neuronal phase-response curves of modeled
cortical neurons. J Comput Neurosci 2008, 26:289-301.
P305
Modelling impairment of evoked gamma range oscillations in
schizophrenia
Christoph Metzner1*, Achim Schweikard1, Bartosz Zurowski2
1
Institute for Robotics and Cognitive Systems, University of Luebeck, 23538
Luebeck, Germany; 2Department of Psychiatry, University of Luebeck,
Schleswig-Holstein, 23538 Luebeck, Germany
E-mail: metzner@rob.uni-luebeck.de
BMC Neuroscience 2015, 16(Suppl 1):P305
Abnormal oscillatory activity in schizophrenia has been found in a wide
range of experimental paradigms [1]. For example, schizophrenic patients
show reduced evoked gamma activity, which has been associated with
negative symptoms, and increased spontaneous gamma activity, which
has been associated with positive symptoms [2]. However, the underlying
mechanisms remain elusive. Here we investigated the impact of circuit
abnormalities on oscillatory activity in the gamma range (> 30 Hz) by
simulating auditory entrainment in an established computational model of
the primary auditory cortex [3]. Auditory click entrainment experiments
showed that for schizophrenic patients EEG/MEG power decreased at 40
Hz and increased at 20 Hz in response to 40 Hz drive but no differences
between were visible in response to 30 Hz drive [4,5].
Here we used the primary auditory cortex model from Beeman [3] and
simulated click train stimulation at 40 Hz, to investigate gamma
entrainment deficits, and at 30 Hz as a control condition. Without
alterations the model entrained at the driving frequency of 30 and 40 Hz,
respectively. Similar to previous approaches [6], however, focusing on
evoked rather than spontaneous activity, we next explored the effects of
(1) connectivity disturbances (reduced (a) recurrent excitation, (b)
pyramidal cell input and (c) total connectivity), (2) prolonged GABAergic
decay time constant, and (3) reduced inhibitory output.
All three interventions in connectivity (1a-c) led to an increase in 40 Hz power
for 40 Hz drive, contrary to human EEG/MEG experiments. A prolonged
GABAergic decay time constant produced a reduction of power at 40 Hz and
an increase in power at 20 Hz, for the 40 Hz drive, which concurs with [4,5].
BMC Neuroscience 2015, Volume 16 Suppl 1
http://www.biomedcentral.com/bmcneurosci/supplements/16/S1
Furthermore, for the 30 Hz drive, no differences to the standard model were
observed. Reduction of inhibitory output led to decreases in power at 40 Hz
for 40 Hz drive but no increases at 20 Hz. In the 30 Hz drive condition, a
decrease was visible, in contrast to experimental data [4,5].
In conclusion, only prolonged GABAergic decay time constants (2), but not
interventions (1) and (3) led to changes in entrainment comparable to
experimental evidence in agreement with previous modeling approaches [5].
Our simulations suggest that prolonged time constants at GABAergic synapses
might play a key role in abnormal evoked gamma rhythms in schizophrenia.
However, since we only investigated one intervention at a time, further studies
are needed to investigate the complex interactions of these circuit
abnormalities. Furthermore, it remains unclear if the same mechanism also
underlies increased spontaneous gamma activity in schizophrenia.
References
1. Gonzalez-Burgos G, Lewis DA: GABA neurons and the mechanisms of
network oscillations: implications for understanding cortical dysfunction
in schizophrenia. Schizophrenia Bulletin 2008, 34(5):944-961.
2. Gordon E, Williams L, Haig AR, Wright J, Meares RA: Symptom profile and
gamma processing in schizophrenia. Cognitive Neuropsychiatry 2001,
6:7-19,2001.
Page 200 of 200
3. Beeman D: A modeling study of cortical waves in primary auditory
cortex. BMC Neuroscience 2013, 14(Suppl 1):P23.
4. Kwon JS, ODonnell BF, Wallenstein GV, Greene RW, Hirayasu Y, Nestor PG,
Hasselmo ME, Potts GF, Shenton ME, McCarley RW.: Gamma frequency-
range abnormalities to auditory stimulation in schizophrenia. Archives of
General Psychiatry 1999, 56(11):1001-1005.
5. Vierling-Claassen D, Siekmeier P, Stufflebeam S, Kopell N: Modeling gaba
alterations in schizophrenia: a link between impaired inhibition and
altered gamma and beta range auditory entrainment. Journal of
Neurophysiology 2008, 99(5):2656-2671.
6. Spencer KM: The functional consequences of cortical circuit
abnormalities on gamma oscillations in schizophrenia: insights from
computational modeling. Frontiers in Human Neuroscience 2009, 3.
Cite abstracts in this supplement using the relevant abstract number,
e.g.: Metzner et al.: Modelling impairment of evoked gamma range
oscillations in schizophrenia. BMC Neuroscience 2015, 16(Suppl 1):P305