Systems Biology and Metabolic Engineering in Bacteria - Systems Biology of Metabolic and Signaling Networks - page 352

simulation (MASS) models. Enzymes, and their changes in activity, are explicitly
represented in the model. However, the model remains a linear model and the
description of regulatory effects is limited to small perturbations of the steady state.
Such a description will never be capable of describing transient behavior or
oscillations. Another promising formalism, termed IOMA (integrated omics meta-
bolic analysis) (Yizhak et al. 2010 ), uses the stoichiometry matrix as in FBA, but
complements the description by Michaelis-Menten type kinetic rate equations. The
model predictions are compared to proteomic and metabolomic data, and the
optimal solution is obtained by quadratic programming. This methodology
compares very well with other available algorithms for standard data sets in E. coli .
These success stories are not limited to E. coli . Szappanos et al. ( 2011 ) built an
integrated model of yeast metabolism, including regulatory interactions, by quanti-
tatively measuring interactions between more than 180,000 gene pairs encoding
metabolic enzymes. They combined the regulatory model with the established
metabolic network and developed a machine learning algorithm for comparing
experimental data with model predictions. See Gerosa and Sauer ( 2011 ) and Reaves
and Rabinowitz ( 2011 ) for recent reviews on integration of metabolism with
different kinds of cellular regulatory mechanisms.
Enormous progress has been made with integrated models for bacteria with
small genomes, in particular Mycoplasma species. These bacteria have a greatly
reduced genome, containing only between 500 and 700 genes. Because they live in
a relatively constant environment, their metabolism is simpler than that of larger
bacteria such as E. coli or Bacillus subtilis . The integration of experimental data
with a metabolic model of Mycoplasma pneumoniae has shown remarkable predic-
tive power (Yus et al. 2009 ). The whole genome, integrated model of Mycoplasma
genitalium , a human urogenital parasite containing only 525 genes, has recently
been completed (Karr et al. 2012b ). The same group has also developed a generic
tool for assembling such models (Karr et al. 2012a ). Even though the task is
simplified by the size of the genome, eventually, models of comparable detail and
predictive power will certainly become available for larger bacteria. The conceptual
tools are largely in place.
13.6 Applications: Modifying Existing Networks and De
Novo Design of Metabolic Pathways
As shown already by several examples above, modeling metabolic networks of
microorganisms, and integrating these networks with regulatory interactions in the
cell has led to new, fundamental functional mechanistic insights in these organisms.
In addition, the knowledge can be used to rationally modify existing networks, or
design networks de novo, for biotechnological applications (Oberhardt et al. 2009 ).
As early as 2003, an algorithm called OptKnock for determining optimal gene
knockouts to improve specific metabolic functions was developed and successfully