Biology Reference
In-Depth Information
have seen, they decompose the system into modules. Structuring according to
functional criteria distorts the picture anyway, and structuring into 'unbiased'
modules still treats a nearly decomposable system as a completely decomposable
one and therefore represents the interactions of the modules in an idealized
way. We conclude that the systeomic models envisaged by Kitano are top-down
models that distinguish themselves from the models of all classes described so
far by the perspective of not simplifying the system.
It is noteworthy that simplification by data reduction, though it looks like a
deficiency, is usually intended. One reason is that models that involve too many
variables become easily intractable, not only analytically but also numerically,
so that they cannot even be used for running simulations. There is another reason
for simplification: We can learn more about the main processes within a system if
we look at them without being influenced by too many minor sideways that only
blur the picture of the core processes. Models have to be simplified or idealized to
be informative (Wimsatt, 1987 and in press; Wolkenhauer & Ullah, this volume).
A simplified model that still reproduces the relevant part of the dynamics of the
modeled system can explain the principles of the dynamics, since one can read
from the models how the dynamics of the biological systems is brought about.
The (fictitious) 'realistic' model, in contrast, simply reproduces the dynamics.
It allows the faithful simulation of all processes of the considered kinds that
take place in an organism. It has predictive power, but little or no explanatory
value (see, e.g., Cartwright, 1989; Lewontin, 1997). Its main value will be as an
easily tractable analogue of organismic processes on which virtual experiments
may be performed. The processes of such
in silico
experiments are measurable
without technical restrictions that hold with respect to the natural systems, but the
dynamics of the simulation demands itself for explanation by simplifying models
(Krohs, 2006b). Anyhow, data about the biological system may be obtained
vicariously from the simulation. This follows a long and interesting tradition that
also has learned about limits of such a transfer (Morgan, 2003). As it is highly
unlikely that a nonmaterial system could reproduce the dynamics of a material
system exactly (Griesemer, 2005), and since 'omic' data sets provide dynamical
data of poor resolution anyway, it seems reasonable that Kitano does not aim
for better than 20% error.
20
The systeome project aims to collect data without providing a strategy to arrive
at explanatory models. Though coming under the label of systems biology, it
turns out to be a purely 'omic' project, as is also made clear in its name. The only
improvement with respect to other 'omic' projects is that it integrates a dynamic
perspective, but instead of taking explanatory advantage from this perspective as
20
Although this might not be relevant to the success of the systeome project, it would be interesting to learn
how many of the results concerning human and mouse systeomes can be expected to be distinguishable within
this error margin.
