Genome-Scale Models of Metabolic and Regulatory Networks - Systems Biology: Networks, Models, and Applications - page 245

number of reactions that are allowed to carry flux is minimized while
still requiring that a reasonable level of biomass production is
sustained. Upon application of this method to a genome-scale E. coli
metabolic network it was found that only 31% and 17% of the reactions
in the model were required to support growth on glucose minimal and
rich medium, respectively. This high degree of redundancy implies
robustness of the metabolic network function against both environ-
mental and genetic changes.
Genome-scale models can also be used to obtain insights into
the experimentally observed apparent nonessentiality of a large frac-
tion of genes in many organisms under standard laboratory conditions
[66,67]. In a recent study, FBA predictions of metabolic gene deletion
strain growth rates under multiple media conditions in yeast were
used to identify reasons for the observed in vivo dispensability [68].
The analysis indicated that the dominant explanation for apparent
dispensability were condition-specific roles for metabolic genes. Much
fewer dispensable genes were buffered by a duplicate gene capable
of acting as an isozyme and only a small fraction of genes were actu-
ally buffered by metabolic network flux reorganization. These results
indicate that while the metabolic network functionality is in general
robust to most gene deletions, the majority of metabolic genes are still
necessary to allow growth in a wide range of nutritional environments.
GLOBAL ORGANIZATION OF METABOLIC NETWORKS
In addition to the robustness and redundancy questions discussed
above, genome-scale models have also been used to analyze more
generally the functional organization of metabolic networks. The
most common type of analysis of this type is the identification of
functional modules in biological networks. The modularization of
biochemical networks has been identified as one of the central goals
for analyzing the behavior and function of these networks [69].
Correlated reaction sets (Co-Sets) contain reactions that are function-
ally connected in all possible functional states of the network under
a particular environmental condition [70]. The difference between
modules defined using analysis of the stoichiometric matrix and
modules defined, for example, using graph theoretic methods is
that the former definition accounts correctly for all the functional
connections implicit in the network whereas the latter definitions
may ignore important connections due to, for instance, cofactor
balancing.
Pathway-based methods can be used to enumerate Co-Sets by
identifying reactions that are co-utilized in all extreme pathways or
elementary flux modes. Co-Sets were calculated with extreme pathway
analysis of the metabolic networks of H. pylori and H. influenzae [61,64].
Nonobvious sets were identified, including reactions involved in