ANIMAL FEED
HENS ABB.
OGY
ELSEVIER Animal Feed Science Technology 64 (1996) 77-89
Multiphasic analysis of gas production kinctics for
in vitro fermentation of ruminant feeds
Jeroen C.J. Groot **, John W. Cone °, Barbara A. Williams ‘,
Filip M.A. Debersaques 4, Egbert A. Lantinga *
* C.F. de Wit Graduate School Production Ecology (PE), Depariment of Agronomy. Wageningen Agricultural
University. P.O. Box 341. 6700 AH Wageningen. The Netherlands
DLO- Institute for Animal ‘and Health (1D-DLO). Depariment of Ruminant Nutrition, Lelystad,
The Netherlands
© Wageningen Institare for Anita Sciences (WIAS), Deparoncnt of Animal Nuirition, Wageningen
Agricultural University, Wageningen, The Netherlands
4 Polytechnic-CFL, Gent, Belgium
© CT. de Wit Graduate School Production Ecology (PE), Depariment of Theoretical Production Ecology.
Wageningen Agricultural University, Wageningen, The Netherlands
Accepted 28 February 1996
Abstract
Recently developed time-related gas production techniques to quantify the kinetics of ruminant
feed fermentation have a high resolution. Consequently, fermentation processes with clearly
contrasting gas production kinetics can be identified. Parameterization of the separate processes is
possible with a suitable multiphasie model and modelling method. A flexible, empirical, multipha-
sic model was proposed for parameterization of gas production profiles. This equation was fitted,
using a two-step method, to four gas production profiles and the maximum fractional rate of
substrate digestion (R4) was calculated for each phase. In the first step, the number of phases and
starting values for parameters of the multiphasic mode! were derived from a combination of either
(1) the measured cumulative gas production profile and its first (rate of gas production) and
second (change in rate of gas production) derivative or (2) @ fitted monophasic curve and the
‘residuals herween this curve and the ohcervations. In the secand step, the starting values were iced
16 fit di- or triphasic models. Using Approach I was complicated by frequent Muctuations in the
gas production rate. However, the combined graph of the fitted monophasic curve and residuals
(Approach 2) gave good results for every profile. The multiphasic model fitted the profiles well.
‘The robustness of the model declined when the number of phases was increased, underlining the
importance of accurate estimations of starting values. It is argued that the model parameters and
* Corresponding author.
10377-8401 /96 /S15:00 Copyright © 1996 Elsevier Science B.V. All rights reserved
PIF S0377-8401(96)01012-7 B J.C. Groot etal. / Animal Feed Science Technology 64 (1996) 77-89
the caleulated Ry are useful for feed quality evaluation. The reoults chow that mathematical
description of gas production profiles requires a two-step approach and a multiphasic model.
Kepronds: Fermentation; Gao production; Multiphasie model; Curve fitting
1. Introduction
In vitro and in sim: methods are used to estimate the organic matter degradability of
ruminant feedstuffs (e.g. Erwin and Elliston, 1959; Tilley and Terry, 1963; Goering and
Van Soest, 1970; @rskov et al., 1980). These methods involve the measurement of
disappearance of substrate during incubation in the rumen (in situ) or in buffered rumen
fluid (in vitro) and are mostly endpoint measurements, The kinetics of degradation can
be determined from differences in the residues after various incubation times. However,
this approach is laborious, expensive and not always accurate enough. Moreover, the
course of fermentation of soluble substrate components cannot be quantified (Pell and
Schofield, 1993).
Alternatively, the kinetics of feedstuff degradation can be determined from the
fermentative gas production, which measures the amount of gas released directly as a
product of the fermentation and indirectly from buffered rumen fluid (Beuvink and
Spoelsira, 1992; Pell and Schofield, 1993; Cone et al., 1994; Theodorou et al., 1994),
‘The kinctics of gas production arc dependent on a scqucnce of proccsses. Upon
incubation, substrates are partly solubilized, The soluble components are rapidly fer-
mentable after incubation. A gradual shift subsequently occurs towards the fermentation
of the insoluble parts, which need to be hydrated and colonized by rumen microorgan-
isms before they can be fermented (Van Milgen et al., 1993). The rates at which these
Processes can take place depend on the initial concentration and composition of the
microbial population, and on the ability of the microorganisms present in the mixture to
colonize, ferment and utilize the fermentation products for growth (Demeyer, 1981;
Hidayat ct al,, 1993). Substrates and their components can be resistant to these processes
to different extents, resulting in substantially different gas production profiles.
Mathematical description of gas production profiles allows analysis of data and
comparison of substrates or fermentation environment characteristics, and can provide
useful information concerning the substrate composition and the fermentability of
soluble and slowly fermentable components of the substrate. Various madels are
available for parameterization of the degradation or fermentation profiles. With early gas
production techniques, still being used, with large glass syringes (e.g. Hidayat et al.,
1993, Khazaal ‘and @iskov, 1994) or a manomeuic system (c.g. Garg and Gupta, 1992,
Waghor and Stafford, 1993), only rather simple gas production profiles can be
obtained, which can be described with rather simple models. These models are often
based on first-order kinetics, assuming a constant fractional rate of fermentation (e.g.
@rskov and McDonald, 1979).
More recent gas production equipment (e.g. Pell and Schofield. 1993: Cone et al.
1994; Theodorou et al., 1994) supplies much more accurate gas production profiles, in
which different underlying processes can be identified, necessitating the use of a more JCJ, Groot et al. / Animat Feed Science Technology 64 (1996) 77-89 7”
sophisticated model. France ct al. (1993) postulated an empirical time-dependent
increase of the fractional rate, In the models of Schofield et al. (1994) the fractional rate
of fermentation increases to a maximum, dependent on the gas produced in a sigmoidal
fashion, following logistic or Gompertz behaviour. Considering the differences between
gas production profiles and the possibilities offered by the high-resolution measurements
using madem equipment. a more flexible model is required. In this paper. we present a
new and flexible model for gas production profile parameterization and a flexible
approach to quantify the number of phases in such profiles and to obtain estimations of
the parameter values, to be used as starting values im a computer curvefit program. The
model and the modelling approach are tested for a range of profiles.
2. Materials and methods
2.1, Model
A sigmoidal equation was adopted to describe the kinetics of the separate phase in the
‘gas production profiles:
6=E— Q
In this equation, G (mlg~' OM) denotes the amount of gas produced per gram of
‘organic matter (OM) incubated, at time r after incubation. A, (ml g™'! OM) represents
the asymptotic gas production. B, (h) is the time after incubation at which half of the
asymptotic amount of gas has been formed, and C; is a constant determining the
sharpness of the switching characteristic of the profile. The value of i indicates the
number of phases in the profile (i = 1,n).
‘The curvature is determined by B and C, resulting in a high flexibility. C determines
the curvature and thereby the position of the point of inflection. The impact of C is
shown in Fig. l(a), For C-< |. the profile has no point of inflection (r > 0), As the value
‘of C increases, the profile becomes sigmoidal, with increasing slope, until for C > « 4
step function is obtained, with an initial slope of zero. The following equation gives the
solution to the second derivative cqual te zero, ic. the point of inflection (,):
cH1yve
1-8 ( ary } (2)
Fig. 1(b) illustrates the effect of increasing C on the relation between f, and B,
expressed as the ratio 4,/B.
From the parameters B and C, the fractional rate of substrate digestion (R) can be
calculated, if a fixed linear relationship is assumed to exist between substrate fermenta-
tion and gas production. The proportion of undigested, potentially digestible substrate is
given by P(t) =[A,—G\(1)]/A,. Thus, R can be calculated as in the equation
lap cre! 5
BB ar” Bee (9)
for C>1. R increases to reach a maximum (Rj), when the size of the microbial 80 JCJ. Groot et al, / Animal Feed Seience Technology 64(1996) 77-89
(2)
——— «0
220
020
a0
Fig. 1. Characteristics of the model. (a) The sigmoidal curve described by Eq, (1), with A= 100, = 15 and
C= 1 (continuous line), 3 (dashed line), 6 (dotted line) and 9 (dot—dashed line), #@, Point of inflection of the
curves, adjusted after Thomley and Johnson (1991), (b) The ratio between the point of inflection (,) and
half-time (B), as affected by the value of C (Eq. (20). (C) The relative rate of fermentation (Eg. (3)) for
ues B = 15 and C= 1.5 (continuous line), 3 (dashed fine), 6 (dotted line) and 9 (dot-dashed
int of maximum fractional rate (re,,. Ry) of the curves.
population no longer limits the fermentation of the particular feedstuff component (Fig.
1(c)). The time after the start of the incubation at which Ry occurs, it. tp,,. solved
from dR/dr=0;
fp, = B(C- 1)" co)
characterizes the rate of growth of microorganisms and colonization of the feed
component. Ry, is reached when the microbial population no longer limits fermentation,
and digestion is not hampered by chemical or structural barriers imposed by the
feedstuff component at this point,
Gas production profiles can be described with the model using non-linear curve-fit-
ting programs (¢.g. NLREG, Sherrod, 1995). However, to do this, the number of phases
has to be estimated, To limit the number of calculations and calculation time, estimates
of the parameter values can be made and used as starting values.
2.2. Estimation of the number of phases and parameter values
2.2.1. Approach 1: cumulative gas production and first and second derivative
Gas production profiles reflect a concatenation of fermentation processes. Throughout
the incubation, these processes either start at their maximum rate and subsequently
decline, or the rate increases to a maximum and declines thereafter. This allows the
separation of processcs which differ in their maximum fermentation rate and in the time
at which this maximum rate is reached. Therefore, characteristics of the different phases
in a profile could be deduced from changes in the measured rate of fermentation, using
the first and second derivative. Here, the observed smoothed rate and change in rate of
gas production will be used, calculated at time ¢ as the average of the values at ¢— 1, ¢
and t+.
Mathematically, maxima in the first derivative and the solution of the second
derivative equals zero represent a point of inflection (1,). This #, can be used as an JCJ. Groot et at. / Animat Feed Science Technology 64.(1996) 77-89 81
estimate of the time after incubation at which half of the asymptotic amount of gas has
been formed (see Fig. (a) and Fig. 1(b)). This value is used as a starting value for B, in
the multiphasic curve fitting, The starting value of A, can be estimated as twice the gas
production at B,, subtracting the gas production from previous phases.
2.2.2. Approach 2: fitted monophasic model and residuals
Differences between phases are often obscured by limited differences in the fermenta-
tion rates of successive phases. An alternative approach is to fit a monophasic model to
a gas production profile with different phases. This will result in alternating periods of
overestimation and underestimation of the gas production of the separate phases, and
will lead to a systematic pattem in the plot of residuals, from which the number of
phases can be determined, as will be demonstrated below.
In non-linear curve fitting, the residuals are minimized (Motulsky and Ransnas,
1987), so that deviations of observations above and below the fitted monophasic model
will be similar, If the phases are: symmetrical (¢— B), the monophasic model will
intersect the observations at half-time and the plot of residuals intersects with the x-axis.
This value 1s used as the starting value tor #, in the multiphasic curve fitting. From the
combined plot of the monophasic model and the residuals, the starting value of A, can
be estimated from the figure as twice the gas production at B,. corrected for gas
production in previous phases.
2.3. Curve fitting of multiphasic models
Eq. (1) is used in the second step, in which i equals the number of phases
distinguished in the first step. From the first step, the starting values for the non-linear
curve fitting are also estimated as described above. Experimental fermentation profiles
of various substrates were fitted to assess the applicahility of the two-step method and
the multiphasic mode]. The NLREG package (Sherrod, 1995) was used. to carry out the
non-linear regression analyses.
‘The significance of the phases was tested. Phases were added w dre wodel, if he
addition of the extra parameters. resulted in a significant improvement of the fit (F-test,
P<0.001, see Motulsky and Ransnas (1987) and the parameters were significantly
different from zero (t-test, P < 0.001; sce: also Zwietering et al. (1990)).
The Durbin—Watson coefficient for autocorrelation was used to test if the model
approaches the data (pnodness af fit), withont systematic errors. Significant autocorrela.
tion indicates either that the model lacks one or more key variables or that the model is
inappropriate for the data values (Neter and Wasserman, 1974).
The robustness, ie. We seusilivily Ww differemt starting values, of the multiphase
model was tested by using deviating starting values (up to 100% from the values
estimated in Step 1) for the non-linear curve fitting procedure.
2.4. Gas production profiles
The gas production profiles of four very different substrates were used to test the
model. These profiles differed owing to the substrates used and the method of incuba- 82 JCJ. Groot et al. / Animal Feed Science Technology 64 (1996) 77-89
Table |
‘The concentration of organic matter (OM), crude protein (CP), neutral detergent fibre (NDF), acid detergent
fibre (ADF) and acid detergent lignin (ADL) in the substrates (gkg~'DM) used in the gas production
experiments
Substrate OM cr NDF ADF
Rye NDF 938 a 915 $36
Sorghum 829 45 663 376
Grass 887 170 538, 251
Clover 871 267 260 230
tion, and were selected to demonstrate the wide applicability of the presented model,
‘The substrates were isolated NDF (neutral detergent fibre) from a young regrowth of
Italian ryegrass (oven-dried at 70°C), mature sorghum stems (air-dried), whole crop
ryegrass regrowth (air-dried) and red clover (freeze-dried), subsequently referred to as
rye NDF, sorghum, grass and clover. All samples were ground to pass a 1 mm screen.
‘The chemical composition of the substrates is given in Table 1.
‘The grass and clover samples were incubated using the method described by Cone et
al. (1994), characterized by a high concentration of inoculum (1:2, v/v). The medium
used in dis meted is based om dat described by Menke ct al. (1979) and adjusted by
Steingass (1983), and contains minerals, reducing agent and buffer. Gas production was
recorded using an automated system.
‘The rye NDF and sorghum were tested for gas production according ta Theodorau et
al. (1994), A low inoculum concentration (1:30, v/v) was used. The medium contained
minerals, volatile fatty acids, vitamins, protein, reducing agent and buffer. Gas produc-
tion was measured manually.
3. Results and discussion
3.1. Applicability
‘The gas production profiles of the four samples tested differed greatly in their
curvature and in the total amount of gas produced (mlg~'OM). The gas profiles for
grass and claver showed rapid gas production in the early stages of fermentation, and the
time after which gas production had virtually stopped was shorter than for the other
substrates; this was also due to the incubation of these samples in poorly diluted rumen
fluid.
From the intersection of the descending part of the curve of the calculated second
derivative with the x-axis (Approach 1), a first phase of each profile could be identified
(Fig. 2) and estimates of the parameter values could be made (Table 2). However,
frequent fluctations occurred in the calculated second derivative after 20h in the rye
NDF and sorghum profiles. and after 10h in the grass and clover profiles. Therefore.
‘only intersections of the descending second derivative with the x-axis were used as an
indication for a phase after systematic positive values (two or moré observations). J.C4. Groor et al. / Animal Feed Science Technology 64 (1996) 77-89 83
Hereby, a second phase was found in the sorghum, grass and ¢lover profiles and a third
phase in the grass and clover profiles (Fig. 2; Table 2),
The use of a monophasic model (Approach 2) was complicated by the high rate of
gas production during a short period immediately following incubation in the grass. and
clover profiles, resulting in underestimation by the monophasic model, and thus positive
residuals. which was not the case with the rye NDF and sorghum profiles. Here, the end
of the first phase was estimated from the first minimum in the plot of residuals and
parameter values were estimated at this point (2B,, A; Fig. 3(c) and Fig. 3(d)). The
systematic Mucwations im the residuals between dhe fitted monophasic curve and the
observations resulted in the identification of two phases in the rye NDF and sorghum
profiles and three phases in the grass and clover profiles (Fig. 3; Table 2).
The estimations for the number of phases in the profiles and the starting values were
compared with results of multiphasic non-linear regression analysis on the profiles.
Phases were added to the model until no significant improvement of the fit or parameter
significance could be reached. The final number of phases (7) in the model leading to
the last significant improvement of the F-value (P < 0.001) and significant parameters
Gas production imi/a OM) 18 2" Derivative:
oT (a) _ (b) 2”
~t «
Time th)
Pig. 2. Observed gas production prone (O), with frst (©) and second (v7) denwvative, Ine dashed ines
indicate estimates of (B,, 0.5A,),(B;, A, +0.5A,) and (B, A, + A, +0.54,). (a) Rye NDF; (b) sorghum;
{c) grass; (d) clover. 84 JCJ, Groot et al, / Animal Feed Science Technology 64 (1996) 77-89
(#test; P < 0.001) amounted to two for the ryc NDF and sorghum profiles and three for
the grass and clover profiles (Fig. 4; Table 2). The values of A, and 2, had been
estimated accurately using both Approach | and Approach 2, except for A, and A, in
the grass. profile, owing to an underestimation of the B, in this profile. These results
indicate the wide applicability of the two-step method, using Approach 2 in Step 1 and
the multiphasic model for parameterization of gas production profiles in Step 2.
3.2. Goodness of fit
The Durbin-Watson coefficients indicated significant auto-correlation only for the
sorghum profile (Table 2). The significant autocorrelation for the multiphasic fit of the
sorghum profile was probably due to deviation of the first observation from the curve.
This observation may have represented a small first phase, but could not be fitted as
such, because the number of observations in the first hours of the incubation period was
too low,
3.3. Robustness
Changing the starting values of parameters had no effect on the outcome of the
monophasic curve fitting, although there were differences in the number of iterations
before convergence was reached. The sensitivity of the curve fitting procedure to
changes in the starting values increased as the number of phases in the model increased.
When starting values that deviated considerably (greater than 25%) from those obtained
Table 2
Estimated (according 10 Approeches | and 2: Est. 1 and Est. 2, respectively) and calculated (Fit) values of
parameters A B (h) and C (-) in di- and triphase models
Substrate Phase A(mig~!OM) Boh) c F-Test DW te Rue
Est Eat 2 Fit 7, Esl Es? Fit 7, Te (pw)
RyeNDF 1 390 320 313 33814 Id 14 701 352 388 2629 146 18.0 0.140
2 ome 1 Ue Wine 39 We 3 3 90 (1.10) 619 wust
Sorghum 1124 126 194 12 11 12-21.0 230 16.5 1008 108 13.5 0.098
2 122: 1S2 151 183 49 48 $2 69.2 3.73 16.4 (1.18) 71.7 0,038
Gus 1 70 @ 4h OR E242 LD 49 142 187 OL Le
2 3% 104 152 202 6 8 9 668 240 213 (@.98) 10.2 0.137
3 728 102 «52:72:35 «37 «(37 344 5.26 73 48.2 0.088
Clee 1 100 9142 4 2 2 76 110 17.2 448 12s 02 OS87
2 92 97 «(120 21.7 10 > 9 187.8 3.09 33.6 (LIS) 111 0188
3 4 BITS 26 1984 138158 30.9 0.414
The F-test indieales the itmprovemeat of the Mt after adding the last significant (P
Vous aimerez peut-être aussi
- Segmento de ClientesDocument5 pagesSegmento de ClientesBel Minda CostaPas encore d'évaluation
- Hoja de Trabajo 2 - FODA Personal - FisicoDocument1 pageHoja de Trabajo 2 - FODA Personal - FisicoBel Minda CostaPas encore d'évaluation
- 2 Introducción A La MicrobiologíaDocument30 pages2 Introducción A La MicrobiologíaBel Minda CostaPas encore d'évaluation
- Segmento de ClientesDocument5 pagesSegmento de ClientesBel Minda CostaPas encore d'évaluation
- 1 IntroDocument44 pages1 IntroBel Minda CostaPas encore d'évaluation
- Caja de Herramientas - MODULO Disenio Estrategico - Taller 1 y 2Document20 pagesCaja de Herramientas - MODULO Disenio Estrategico - Taller 1 y 2Bel Minda CostaPas encore d'évaluation
- TesisDocument25 pagesTesisBel Minda CostaPas encore d'évaluation
- Los Bovinos UnidadDocument6 pagesLos Bovinos UnidadBel Minda CostaPas encore d'évaluation
- Formato de Resumen EjecutivoDocument5 pagesFormato de Resumen EjecutivoBel Minda CostaPas encore d'évaluation
- Funciones PlacentaDocument13 pagesFunciones PlacentaBel Minda CostaPas encore d'évaluation
- Funciones PlacentaDocument13 pagesFunciones PlacentaBel Minda CostaPas encore d'évaluation
- Las Botellas de Plastico Son Proyecto 2Document10 pagesLas Botellas de Plastico Son Proyecto 2Raguil RaguilPas encore d'évaluation
- InterculturalidadDocument1 pageInterculturalidadBel Minda CostaPas encore d'évaluation
- Novedades Veterinaria 2017Document16 pagesNovedades Veterinaria 2017Bel Minda CostaPas encore d'évaluation
- Del Poliéster A La Forma Escultórica - Rachel Whiteread y Duane HansonDocument11 pagesDel Poliéster A La Forma Escultórica - Rachel Whiteread y Duane HansonBel Minda CostaPas encore d'évaluation
- Medellin PDFDocument62 pagesMedellin PDFClaudia patricia menozaPas encore d'évaluation
- Efecto Del Tipo de DietaDocument41 pagesEfecto Del Tipo de DietaBel Minda CostaPas encore d'évaluation
- TMM Tema4 Video Digital PDFDocument116 pagesTMM Tema4 Video Digital PDFMimisita TrujiDaviPas encore d'évaluation
- Guía Biología C - Octubre 2016 - Versión ColorDocument139 pagesGuía Biología C - Octubre 2016 - Versión ColorBel Minda CostaPas encore d'évaluation
- Hervás - 2001 (Tesis Doctoral) PDFDocument249 pagesHervás - 2001 (Tesis Doctoral) PDFJuanPabloSalazarPas encore d'évaluation
- TFG - Materiales y Técnica Empleados Durante El Moldeado y CopDocument37 pagesTFG - Materiales y Técnica Empleados Durante El Moldeado y Copben7Pas encore d'évaluation
- Plasticos PDFDocument15 pagesPlasticos PDFfcozambranaPas encore d'évaluation
- Guion Tecnico 2013 PDFDocument24 pagesGuion Tecnico 2013 PDFKOKYLOPas encore d'évaluation
- Tropa y Marinería - Ejemplo Test Razonamiento EspacialDocument8 pagesTropa y Marinería - Ejemplo Test Razonamiento EspacialTekas2014Pas encore d'évaluation
- Del Poliéster A La Forma Escultórica - Rachel Whiteread y Duane HansonDocument11 pagesDel Poliéster A La Forma Escultórica - Rachel Whiteread y Duane HansonBel Minda CostaPas encore d'évaluation
- M3a1bd1 - Guía FDocument2 pagesM3a1bd1 - Guía FBel Minda CostaPas encore d'évaluation
- Guia Convivencia Cast PDFDocument30 pagesGuia Convivencia Cast PDFZirethPas encore d'évaluation
- Capitulo 7 - Estrategias DidacticasDocument26 pagesCapitulo 7 - Estrategias Didacticasapi-3697492100% (2)
- SEGUNDOCICLODocument18 pagesSEGUNDOCICLOLidia MirandaPas encore d'évaluation
