Académique Documents
Professionnel Documents
Culture Documents
201000163
Abstract: Ayurveda, the traditional Indian medicine is one examples where the traditional medicinal use of the plant
of the most ancient, yet living medicinal traditions. In the matches with the medicinal use of the drug that is structur-
present work, we developed an in silico library of natural ally similar to a plant component. With this approach, we
products from Ayurveda medicine, coupled with structural have brought to light a number of obscure compounds of
information, plant origin and traditional therapeutic use. natural origin (e.g. kanugin, norruffscine, isoazadirolide)
Following this, we compared their structures with those of that could provide the basis and inspiration for further lead
drugs from DrugBank and we constructed a structural simi- development. Apart from the identification of novel natural
larity network. Information on the traditional therapeutic leads in drug discovery, we envisage that this integrated in
use of the plants was integrated in the network in order to silico ethnopharmacology approach could find applications
provide further evidence for the predicted biologically in the elucidation of the molecular basis of Ayurveda medi-
active natural compounds. We hereby present a number of cine and in drug repurposing.
Keywords: Natural products Ayurveda Structural similarity Traditional medicine Biological activity
1 Introduction
Traditional medicines are used by 60 % of the worlds popu- random screening for every target is impractical, both eco-
lation. The World Health Organization (WHO) lists 21 000 nomically and with respect to time.
plants used for medicinal purposes all over the world. It is, therefore, not surprising that the chemoinformatics
Among these, 2500 species are in India and 150 of them community has recently set focus on the application of
are used commercially on a reasonably large scale, making computational approaches for the identification of bioac-
India the largest producer of medicinal herbs in the tive NPs, which is further demonstrated by the amount of
world.[1] recent review articles in the field.[47]
Traditional medicines have been identified through real Nevertheless, actual in silico research involving NPs is still
life experiences and direct observations in people with dis- rather rare in the scientific literature. Recent studies focus
eases and represent highly complex biological systems. Tra- either on the exploration of the chemical space covered by
ditional Indian Medicine (TIM), also known as Ayurveda, in NPs,[810] or on the application of NP libraries in ligand-
particular, was originated and developed between 2500 based[1114] and target-based[15, 16] virtual screening.
and 500 BC and has accumulated many generations of ob- In Traditional Chinese Medicine (TCM) research in particu-
servations with well-organized and documented data.[2] It lar, a number of studies have recently emerged, which
could, thus, be regarded as an informative repository of apply data-mining[17, 18] and systems biology approaches[19]
clinically proven natural products (NPs) that work effec- to explore the molecular basis of TCM and to relate its ter-
tively and safely.[3] minology with the western medicinal context.
It has long been recognized that NP structures have the A distinct bottleneck in the in silico evaluation of the bio-
characteristics of high-chemical diversity, biochemical spe- logical activity of NPs has been the unavailability of large
cificity and other molecular properties that make them fa-
vorable as lead structures for drug discovery. But even [a] H. Polur, T. Joshi, C. T. Workman, I. Kouskoumvekaki
though it is generally accepted that they contain inherently Department of Systems Biology, Center for Biological Sequence
Analysis, Technical University of Denmark
large-scale structural diversity and could play a protagonist
Kemitorvet, Building 208, DK-2800 Lyngby, Denmark
role for discovering new drugs, assessing this diverse chem- phone/fax: + 45 45256162/ + 45 45931585
ical space efficiently and effectively has remained a chal- *e-mail: irene@cbs.dtu.dk
lenge for medicinal chemists and pharmacologists. Collec- [b] G. Lavekar
tion and isolation of all the available NPs, followed by Senior Consultant (Tech.) National Medicinal Plants Board,
Chandralok Building, 36-Janpath, New Delhi-110001, India
Mol. Inf. 2011, 30, 181 187 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim 181
Full Paper H. Polur et al.
NP databases with structural and biological information, es- Family Welfare. It included 300 plant species with informa-
sential in order to make NPs accessible for virtual screening. tion on botanical name, part used, Sanskrit name, chemical
Only few, commercially available databases exist, the most constituents and traditional medical usage.
comprehensive being the Dictionary of Natural Product Da- As the original data set did not contain structural infor-
tabase (Chapman and Hall), a structure database containing mation, it was first processed with Reflect.[31] Reflect is a
over 226,000 NPs, with no information, however, about consolidated dictionary that links names and synonyms to
their biological target. A number of proprietary databases source data entries and currently contains 4.3 million small
also exist, such as the Chinese Natural Product Database molecules. 1482 compound names from the original Ayur-
developed at the Shanghai Institute of Materia Medica, veda data set were annotated with a CID (Compound ID)
which are focused on TCM and contain structural informa- from Reflect and a final list of 1184 compounds with
tion of plant components and TCM usage described ac- unique CIDs was obtained after removing the duplicates.
cording to TCM principles. The list of CIDs was subsequently used to query PubChem
The current state of NP databases manifests that the and all structures were retrieved in a single SDF. The chemi-
knowledge on traditional medicinal usage has not been cal structures corresponding to 645 chemical names that
fully exploited as yet. In addition to this, it is apparent that had not been tagged by Reflect were manually retrieved
visibility of Ayurvedic medicine remains much lower as from the literature and added to the SDF. The final data set
compared with TCM.[2] consisted of 1829 unique NPs from 295 Ayurveda plants
In the present study, we developed an in silico library of and was imported in MOE. All structures were washed and
NPs from Ayurveda medicine, coupled with structural infor- energy minimized as before, using the MMFF946 force
mation, plant origin and traditional therapeutic use. Subse- field.
quently, we made use of the traditional therapeutic indica-
tion of the plants according to Ayurveda, to support pre-
2.2 Molecular Descriptors and Similarity Evaluation
dicted biological activities of the NPs they consist of, which
were based on structural similarity pairings with drug com- Five descriptors related to drug-likeness were calculated in
pounds. We envisage that this integrated in silico ethno- MOE, namely molecular weight (MW), hydrogen bond
pharmacology approach could be of use in the elucidation donors (HB-donors), hydrogen bond acceptors (HB-acc),
of the molecular basis of the therapeutic activity of Ayurve- number of rings and number of rotatable bonds.
da medicinal plants, drug repurposing, as well as the iden- According to the Similarity Principle, compounds with
tification of novel chemical entities with attractive scaffolds similar chemical structures usually possess similar physico-
for drug discovery. chemical properties and biological activities.[32] This is the
underlying assumption for making predictions for new
compounds with biological activity. The chemical structures
2 Methods and Material were encoded using two types of fingerprints that capture
different levels of similarities between molecules. The 166
2.1 Data Sets
keys MACCS fingerprints encode the presence or absence
The chemical space of Ayurveda NPs with known biological of predefined substructural or functional groups.[33] The
activity was captured by mapping the compounds against Gpi-DAPH3 pharmacophore fingerprints are based on an
ChemProt.[20] ChemProt is a repository of 700,000 unique expansion of the PATTY pharmacophore feature recognition
chemicals with biological activity for more than 30,500 pro- scheme.[34] This scheme assigns one or more pharmaco-
teins, assembled from both proprietary and freely available phore feature types to all the atoms in a molecule based
databases, including ChEMBL (version chembl 05),[21] Bind- on a predefined list of SMARTS queries. The pharmaco-
ingDB,[22] PDSP Ki Database,[23] DrugBank (version 2.5),[24] phore feature types are: hydrogen-bond donor (D), hydro-
PharmGKB,[25] WOMBAT (version 2009), WOMBAT-PK (version gen-bond acceptor (A), polar = donor & acceptor (P) and
2008),[26] PubChem bioassay (2010),[27] CTD (version 2009)[28] hydrophobic (H). In addition, an extra label (p or pi) can be
and STITCH (version STITCH 2.0).[29] added to each feature if the originating atom or group is
The data set of small molecule drugs was downloaded sp2-hybridized or planar for other reasons. Gpi-DAPH3 is
from the DrugBank v. 2.5[24] and consisted of 4887 com- expressed as triplet feature combinations with a graph
pounds, referred to in DrugBank as small molecule struc- based inter-atom distance binning scheme. Both types of
tures. All compounds were extracted in SDF format and fingerprints are implemented in MOE.
were imported in Molecular Operating Environment (MOE, The similarity between two molecules was measured
v.2008.10).[30] All structures were washed, i.e. all ionizable using the Tanimoto coefficient (Tc).[35] Tc is calculated from
groups were coordinated with neutral pH conditions, and the number of bits in common divided by the total
energy minimized using the MMFF946 force field. number of bits in both molecules and ranges between zero
The original Ayurveda data set was provided in a flat file (maximum dissimilarity) and one (maximum similarity). In
format by the Central Council for Research in Ayurveda and the case of MACCS fingerprints we used a cutoff of 0.85 as
Sidha, Department of AYUSH, Indian Ministry of Health and the empirical threshold for correlating structural similarity
182 www.molinf.com 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Mol. Inf. 2011, 30, 181 187
Biologically Active Natural Products fromAyurveda Traditional Medicine
Table 1. Mean and standard deviation values of selected physicochemical properties related to drug-likeness for NPs from TIM and com-
pounds from DrugBank. The considered properties are: molecular weight (MW), hydrogen bond donors (HB-donors), hydrogen bond ac-
ceptors (HB-acc), number of rings and number of rotatable bonds.
Mean Standard deviation
TIM DrugBank TIM DrugBank
MW 379.5 363.8 257.1 307.5
HB-donors 2.6 2.2 4.3 3.8
HB-acc 4.7 3.8 6.8 5.1
# rings 3.4 2.3 2.4 2.0
# rotatable bonds 4.7 6.7 6.4 9.9
Mol. Inf. 2011, 30, 181 187 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.molinf.com 183
Full Paper H. Polur et al.
184 www.molinf.com 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Mol. Inf. 2011, 30, 181 187
Biologically Active Natural Products fromAyurveda Traditional Medicine
Figure 3. NPDrug network generated by using structural similarity information between NPs from Ayurveda medicine (ellipse) and drugs
from DrugBank (diamonds). A link is placed between a NP node and a drug node if the Tanimoto coefficient is higher than 0.85 and 0.50
for MACCS and GpiDAPH3 fingerprints respectively. Examples of NP-Drug sub-networks, where the therapeutic use of the drug matches
with the traditional use of the plant(s) that contain the NP, are magnified.
Mol. Inf. 2011, 30, 181 187 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.molinf.com 185
Full Paper H. Polur et al.
ing that before the advent of high-throughput screening, comparison between NPs from Ayurveda and drugs, we are
more than 80 % of drug substances were NPs or inspired the first to develop a TIM database with structural informa-
by nature and almost half of the drugs approved since tion and of a size that allows its application in virtual
1994 are based on NPs.[7] For the purpose of the structural screening. With this database, we have brought to light a
186 www.molinf.com 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Mol. Inf. 2011, 30, 181 187
Biologically Active Natural Products fromAyurveda Traditional Medicine
number of obscure compounds of natural origin (e.g. kanu- [9] N. Singh, R. Guha, M. A. Giulianotti, C. Pinilla, R. A. Houghten,
gin, norruffscine, isoazadirolide) that could provide the J. L. Medina-Franco, J. Chem. Inform. Model. 2009, 49, 1010
basis and inspiration for novel lead development. 1024.
[10] T. M. Ehrman, D. J. Barlow, P. J. Hylands, J. Chem. Inform. Model.
Besides the identification of novel bioactive NPs, the che- 2007, 47, 2316 2334.
moinformatics-based NP-Drug network that we developed [11] J. M. Rollinger, P. Mocka, C. Zidorn, E. P. Ellmerer, T. Langer, H.
could also be of relevance in the rationalization of Ayurve- Stuppner, Current Drug Discov. Technol. 2005, 2, 185 193.
da medicinal praxis, as well as in drug repurposing, as in [12] A. Drig, I. Kouskoumvekaki, R. M. Vejborg, P. Klemm, Appl. Mi-
the case of podofilox that we describe above. crobiol. Biotechnol. 2010, 87, 309 317.
[13] T. M. Ehrman, D. J. Barlow, P. J. Hylands, J. Chem. Inform. Model.
2007, 47, 264 278.
[14] R. K. Petersen et al., JCAMD, 2010.
5 Abbreviations [15] J. M. Rollinger, A. Hornick, T. Langer, H. Stuppner, H. Prast, J.
Med. Chem. 2004, 47, 6248 6254.
CID Compound ID [16] Z. Nikolovska-Coleska et al., J. Med. Chem. 2004, 47, 2430
HB-acc Hydrogen bond acceptors 2440.
HB-donors Hydrogen bond donors [17] Y. Cheng, Y. Wang, and X. Wang, Comput. Biol. Chem. 2006, 30,
148 154.
MOE Molecular operating environment
[18] Y. Wang, Y. Jin, C. Zhou, H. Qu, and Y. Cheng, Med. Biol. Eng.
MW Molecular weight Comp. 2008, 46, 605 611.
NP Natural product [19] J. Zhao, P. Jiang, W. Zhang, Brief. Bioinform. 2009, 11, 417
Tc Tanimoto coefficient 430.
TCM Traditional chinese medicine [20] O. Taboureau et al. Nucl. Acids Res. 2011, 39, D367 D372.
TIM Traditional indian medicine [21] P. de Matos et al., Nucl. Acids Res. 2009, 38, D249 D254.
WHO World Health Organization [22] T. Liu, Y. Lin, X. Wen, R. N. Jorissen, M. K. Gilson, Nucl. Acids
Res. 2007, 35, D198 D201.
[23] B. Roth, Pharmacol. Therap. 2004, 102, 99 110.
[24] D. S. Wishart et al., Nucl. Acids Res. 2008, 36, D901 D906.
[25] M. Hewett et al., Nucl. Acids Res. 2002, 30, 163 165.
Acknowledgements [26] M. Olahet al., in Chemical Biology: From Small Molecules to Sys-
tems Biology and Drug Design CRds: S. L. Schreiber, T. M.
Kapoor, G. Wess), Wiley-VCH, New York, 2007, pp. 760 786.
I. Kouskoumvekaki acknowledges financial support from [27] D. L. Wheeler et al., Nucl. Acids Res. 2007, 35, D5 D12.
the Research Council for Technology and Production Sciences. [28] A. P. Davis, C. G. Murphy, C. A. Saraceni-Richards, M. C. Rose-
The authors would like to thank the Central Council for Re- nstein, T. C. Wiegers, C. J. Mattingly, Nucl. Acids Res. 2009, 37,
search in Ayurveda and Siddha of the Indian Ministry of D786 D792.
Health & Family Welfare for providing the original Ayurveda [29] M. Kuhn et al., Nucl. Acids Res. 2009, 38, D552-D556.
[30] Molecular Operating Environment, Chemical Computing Group,
data set. The authors would also like to thank Lars Juhl
Montreal, Canada, 2008.
Jensen for providing the beta version of the Reflect server. [31] E. Pafilis et al., Nat. Biotechnol. 2009, 27, 508 510.
[32] R. C. Glen, S. E. Adams, QSAR Comb. Sci. 2006, 25, 1133 1142.
[33] J. L. Durant, B. A. Leland, D. R. Henry, J. G. Nourse, J. Chem.
References Inform. Model. Comp. Sci. 2002, 42, 1273 1280.
[34] B. L. Bush and R. P. Sheridan, J. Chem. Inform. Model. 1993, 33,
[1] M. Modak, P. Dixit, J. Londhe, S. Ghaskadbi, T. Paul A Devasa- 756 762.
gayam, J. Clinic. Biochem. Nutrition 2007, 40, 163 173. [35] N. Kochev, V. Monev, I. Bangov, in: Chemoinformatics (Eds: J.
[2] B. Patwardhan, D. Warude, P. Pushpangadan, N. Bhatt, Evi- Gasteiger, T. Engel), in Wiley-VCH, Weinheim, 2003, pp. 304
dence-Based Complement. Alternat. Med.: eCAM 2005, 2, 465 305.
473. [36] R. Brown, Y. Martin, J. Chem. Inform. Model. 1996, 36, 572
[3] X. Zhou, Y. Li, d X. Chen, J. Mol. Graph. Model. 2010, 29, 38 584.
45. [37] D. P. Zlotos, Mini Rev. Med. Chem. 2005, 5, 595 606.
[4] J. M. Rollinger, T. Langer, H. Stuppner, Curr. Med. Chem. 2006, [38] M. Lohombo-Ekomba, P. N. Okusa, O. Penge, C. Kabongo, M. I.
13, 1491 1507. Choudhary, O. E. Kasende, J. Ethnopharm. 2004, 93, 331 335.
[5] J. M. Rollinger, T. Langer, H. Stuppner, Planta Medica 2006, 72, [39] H. Huddart, K. H. Saad, J. Exper. Biol. 1980, 86, 99 114.
671 678. [40] C. Canel, R. M. Moraes, F. E. Dayan, D. Ferreira, Phytochemistry,
[6] K. Grabowski, K. Baringhaus, G. Schneider, Natural Product Re- 2000, 54, 115 120.
ports 2008, 25, 892 904.
[7] A. L. Harvey, Drug Discov. Today 2008, 13, 894 901. Received: November 8, 2010
[8] J. Rosn, J. Gottfries, S. Muresan, A. Backlund, T. I. Oprea, J. Accepted: December 20, 2010
Med. Chem. 2009, 52, 1953 1962. Published online: February 10, 2011
Mol. Inf. 2011, 30, 181 187 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.molinf.com 187