Agriculture Reference
In-Depth Information
and
in silico
pigs. These results are promis-
ing and suggest that a generic structure of
the covariance matrix of model parameters
can be used to account for variation among
pigs in simulation modelling. This approach
is in line with that of Knap (1995), who
stated that 'Variation should be introduced
as deep in the model as possible: when
basic model variables are varied, all vari-
ables that depend on them will automatic-
ally display covariance'. A future version of
InraPorc could generate a virtual popula-
tion of pigs to evaluate how the population
responds to different management practices
(e.g. change of diets, slaughtering of ani-
mals, carcass payment grids). Because many
of these practices are based on discrete
events, our view is that this can be ad-
dressed best by repeated simulations of a
deterministic model while accounting for
the variation among individuals.
because we felt that the required informa-
tion was not readily available to the user.
Our goal was to capture the maximum infor-
mation (across animals) using a minimum of
model parameters to ensure that the model
would be robust. Allowing users to change
parameters that are now hard-coded in the
model would certainly make the model
more flexible; however, it would also place
an additional challenge or burden on users
to provide that information.
Variation among individuals is inher-
ent to living systems. However, most nutri-
tional growth models are deterministic and
thus ignore this variation. There are two
approaches to dealing with variation among
animals in the practical application of growth
models. Our approach was to account for vari-
ation among animals and to identify opti-
mum management strategies for the popula-
tion. Another approach is that of precision
feeding, which is discussed by Candido
Pomar and colleagues (see Pomar
et al
.,
Chapter
12,
this volume). Both approaches
rely on phenotyping or monitoring of the
animals, and this information has to be trans-
formed into model inputs. We feel strongly
that there is a future for phenotyping ani-
mals in combination with modelling so that
practical management decisions can be ap-
plied to the herd or to individual animals.
However, (real-time) information on indi-
vidual animals is scarce and methods to ac-
quire and to use this information as model
inputs have to be developed.
Conclusions
We have attempted to develop a model and
user-friendly tool that allows the evaluation
of the response of the pig to nutrient supply.
One of the main challenges during model
development was the identification of the
most important model parameters for which
model users could provide relevant informa-
tion. This meant that certain model param-
eters were made constant in the software
References
Bikker, P., Verstegen, M.W.A., Campbell, R.G. and Kemp, B. (1994) Digestible lysine requirement of
gilts with high genetic potential for lean gain, in relation to the level of energy intake.
Journal of
Animal Science
72,
1744-1753.
Bikker, P., Verstegen, M.W.A. and Campbell, R.G. (1996) Performance and body composition of finish-
ing gilts (
45
to
85
kilograms) as affected by energy intake and nutrition in earlier life: II. Protein
and lipid accretion in body components.
Journal of Animal Science
74,
817-826.
Black, J.L., Campbell, R.G., Williams, I.H., James, K.J. and Davies, G.T. (1986) Simulation of energy and
amino acid utilisation in the pig.
Research and Development in Agriculture
3,
121-145.
Brossard, L., Dourmad, J.Y., Rivest, J. and van Milgen, J. (2009) Modelling the variation in performance
of a population of growing pigs as affected by lysine supply and feeding strategy.
Animal
3,
1114-1123.
Conde-Aguilera, J.A., Barea, R., Le Floc'h, N., Lefaucheur, L. and van Milgen, J. (2010) A sulfur amino
acid deficiency affects the amino acid composition of body protein in piglets.
Animal
4,
1349-1358.
