Agriculture Reference
In-Depth Information
classical model. The main challenge and an
extremely time-consuming task has been
the development of the software tool. This
involved identifying the most relevant and
accessible inputs to run the model under a
variety of conditions and providing outputs
that allow the user to analyse and under-
stand the responses predicted by the model.
Since its initial release, various minor and
major software upgrades have been pub-
lished, most of which concerned improving
the user-friendliness of the tool. The soft-
ware can be downloaded from
www.rennes.
inra.fr/inraporc/
and is free for educational
purposes.
The InraPorc model is structurally very
similar to that developed
40
years ago by
Whittemore and Fawcett (1974), where
body weight gain is modelled as a function
of protein and lipid weight gain. Under nu-
tritionally non-limiting conditions, feed in-
take and protein deposition are modelled
independently and are user inputs required
to run the model. Feed intake is represented
as a function of body weight, while protein
deposition is defined as a Gompertz func-
tion, both of which have to be parameter-
ized by the user. The main reason that we
chose to use feed intake and protein depos-
ition as user inputs is that these can be de-
termined relatively easily by the user.
Through its relationship with body water,
there is a strong relationship between pro-
tein deposition and body weight gain. Like-
wise, feed or energy intake can be measured
directly by the user. Energy not used for
protein deposition and maintenance will be
available for lipid deposition, and lipid de-
position is thus considered an energy sink.
The consequence of this approach is that
errors will accumulate in lipid deposition.
Because it is very difficult to estimate lipid
deposition accurately during growth (and
even at slaughter), it is thus very difficult to
evaluate the accuracy of the predicted lipid
deposition (or related traits). This is a gen-
eric problem in growth models and not spe-
cific for the InraPorc model. Feed restric-
tion, amino acid deficiencies and changes
in the maintenance energy requirement can
affect protein deposition, energy partition-
ing and thus growth.
In the next sections, we will describe
how these responses are modelled and how
this has been incorporated into the software
tool. In doing so, we had to decide how
model parameters can be determined. In
certain cases (e.g. defining the phenotypic
feed intake and growth potential of the
animal) the user has to provide the corres-
ponding model parameters. In other cases,
default model parameters are provided that
can be changed by the user if sufficient
information is available to justify a change.
We also decided to hard-code certain param-
eters in the software tool because we felt
that changing these parameters would re-
quire information inaccessible to most users,
or that it may affect model predictions be-
yond our control.
Using the InraPorc Software Tool
To run a simulation for growing pigs, InraPorc
combines information from three different
modules: the animal profile, a feed sequence
plan and a feed rationing plan. The animal
profile describes the phenotypic
ad libitum
feed intake and growth potential of the ani-
mal. To account for differences among ani-
mals, the user has to provide a minimum of
five parameters: the initial body weight, two
parameters to describe the feed intake curve
and two parameters to describe the protein
deposition curve. In the first version of the
software, the user had to obtain these model
parameters by trial and error by confronting
experimental data with model predictions.
Since 2009, we have included an algorithm
that allows the estimation of model param-
eters directly from experimental observations
through a statistical procedure. Because both
feed intake and growth are dynamic, the user
has to provide feed intake and growth data
for at least two different periods to estimate
the parameters for the feed intake and
growth curves. A feed sequence plan de-
fines the different diets that will be used
during a simulation. The diet composition
can be calculated from the composition of
feed ingredients (Sauvant
et al
., 2004), or
can be provided by the user based on the
