Biology Reference
In-Depth Information
and the way we would have thought of constructing a module to carry out the
function.
Sauro (2004) and Deckard & Sauro (2004) have therefore attempted to invert
the problem of recognising the functions implemented by networks of enzymes
by first defining output functions and then, again with genetic algorithms, evolv-
ing enzyme networks in the computer that compute the function. They hope that
this will allow the inference of a similar function for a module in a biological
network if it is observed to contain a similar configuration of components. This
is an ingenious approach, but requires for its success that we can hypothesise
appropriate functions and set up the genetic algorithm to evolve appropriate
circuits.
In this section, I have attempted to show that there could well be limitations
to our ability to understand completely how a cell works, partly because we
may have difficulties in defining the functions being implemented and partly
because we may not recognise the collection of components that implement the
function. This does not necessarily mean that we cannot predict the responses of
a biological system, as we have another type of modelling available that I have
not yet discussed: computer simulation.
5.
IS SIMULATING CELL METABOLISM THE SAME AS
UNDERSTANDING IT?
Although I have presented theoretical analysis of models as a separate issue from
computer simulation, most practitioners of the former also build simulations of
the networks and phenomena they are analysing. There are a number of reasons
for this. First, simulations of a model can replace experimental observations
as source material for developing hypotheses about behaviours of the system.
Next, once a theoretical analysis has been formulated, the simulation can serve
as an illustrative case, especially for the benefit of those biologists who would
rather look at a graph than read an equation. In both these instances, the lack
of experimental noise in the simulated system (at least, unless the simulation
is stochastic) and the freedom to accurately make any desired alterations to the
properties or quantity of a component, irrespective of their practical feasibility,
are distinct advantages. Furthermore, it is usually easy to simulate larger systems
with additional processes beyond those represented in the theoretical analysis,
where they would make the analysis less tractable. In this way, it may be possible
to test whether factors that were omitted from the theoretical analysis are truly
of lesser importance, or, looking at it the other way round, whether the processes
and components included in the theoretical model are sufficient to generate the
main aspects of the behaviour of a more realistic model of the biological system.
