Biology Reference
In-Depth Information
This branch of systems biology is regarded as being predominant in Europe.
Nevertheless, the Japanese and US-American traditions of top-down modeling
(see Westerhoff & Palsson (2004) and Cassman et al. (2005) for the geographic
assignment) until recently also relied mostly on exemplars from the bottom-up
branchwhen presenting convincing results of systems biology (see Kitano, 2002c).
6.2. The second branch of systems biology: Making sense of 'omic'
data by top-down modeling
The second branch of systems biological modeling proceeds in top-down fashion.
The change with respect to the tradition of top-down modeling of cybernetics
and systems analysis seems to be more dramatic than the change within the
bottom-up tradition. As argued above, modern 'omic' projects, in contradistinc-
tion to their predecessors from pre-genomic days, do not allow much biologically
relevant interpretation and are hardly accessible to mathematical modeling of
any physiological interest. This is overcome in top-down systems biology by a
combination of two strategies. The first strategy is to use data that show not only
one state of a network but also its
dynamics
. To do this, multiple 'omic' data
sets of the same sort, but of different states of a cell, must be collected. This
may be done, e.g., in successive physiological states, or at different times during
the development of an organism, or by monitoring 'perturbed' systems (Ideker
et al., 2001). Perturbation may be effected by altering the system's environment
or by changing its components, which in the case of a genetic network means
modifying or deleting genes. Given the availability of high-throughput methods
and their further improvability, this is within the range of feasibility. These data
provide a basis for modeling. Different ways were found to depict a network
in terms of the interaction of its components, which are not further character-
ized in functional terms, and to account for its dynamics, e.g., Difference-Based
Regulation Finding (Onami et al., 2002) and stoichiometric matrices (Palsson,
2006). However, data on the dynamics of a whole network have only lim-
ited explanatory power as long as the network cannot be structured in a way
that makes its internal dynamics understandable. Therefore, the second strategy
involved in the systems biological evaluation of 'omic' data sets is structuring
the network into manageable subnetworks ('modules'). To accomplish this, the
problem must be solved as to how subnetworks that are physiologically relevant
may be singled out.
Systems biologists decompose a network into modules in two different ways.
The first way to decompose a network of 'omic' scale isolates the subnet-
works in terms of knowledge about network capacities and the contributions of
components to these capacities. Such physiological modules are modeled sepa-
rately, often in a bottom-up manner. Reassembling different modules is another
bottom-up step, but may not result in a model of the complete network. The
