Biology Reference
In-Depth Information
Through these approaches, simulation of metabolism and other cellular
networks has proved an important tool in developing our understanding,
especially as biochemical simulations can be developed and solved relatively
rapidly with modern software. However, simulation alone without an analytical
model has some weaknesses. The main one is that even a small model can
have a large number of parameters, for example, in metabolism, these may be
K m s, Hill coefficients, equilibrium constants, and limiting rate values for the
component enzymes. Hence it soon becomes impractical to determine whether a
particular behaviour is robust over the whole feasible range of every parameter,
or whether different behaviours emerge in different regions of parameter
space. This contrasts with an analytical model, where it should usually prove
possible to specify boundary constraints for the observation of a particular
behaviour. Thus simulating a model will generally not contribute as much to
the understanding of a biological process as does an appropriate analytical
model. On the other hand, even in the simplest cell there are thousands of
different interacting components, and it is inconceivable that an analytical
model can be made of the whole of such a complex system. Currently, it is
not possible to simulate a whole cell either, but it is not inconceivable that
this could be done. The advent of high-throughput analytical technologies (the
'-omics') is accelerating the rate of data capture, and consortia are already
forming to build simulations of E. coli (http://www.ecolicommunity.org/),
S. cerevisiae (http://www.siliconcell.net/ysic/), and the hepatocyte
(http://www.bmbf.de/en/1140.php). What understanding would we have
obtained if we were to succeed in producing a silicon replica of a living cell?
In one respect, we would have substituted a highly complex experimental
object with a highly complex simulation, which might not be any easier to
understand. It would still be necessary to develop hypotheses about the factors
and processes that determine the model's behaviour, and then to test those
hypotheses by appropriate interventions. Some of the advantages of simulation
of small models would still be present. The implementation of the interventions
would be much easier in the case of the simulation, and the results might be
more clear-cut, even if they involved such small changes in a concentration or
rate that they would not be detectable experimentally. We would still be able
to test whether the components and processes described in the simulation are
sufficient to account for known behaviour, though tracking down the reasons
for any discrepancies is likely to prove very difficult on account of the size of
the model and its parameter space.
My view is that models of this size will be useful for integrating information
and concepts about the cell's components in order to predict the behaviour,
and that as a result we will be more successful in predicting the impact of
mutations, environmental alterations and drug treatments. Whether we will really
understand why the model has made the predictions it does is a different matter.
Search WWH ::




Custom Search