Genome-Scale Models of Metabolic and Regulatory Networks - Systems Biology: Networks, Models, and Applications - page 238

reduced cost of the linear programming problem [35]. Similar sensitivity
analysis can also be performed using two variable parameters simulta-
neously, for example, the maximal glucose and oxygen uptake rates.
This approach, called phenotype phase plane analysis (PhPP), allows
subdividing the two-dimensional space spanned by the two parame-
ters into linearly constrained regions, based on the shadow price
structure of the corresponding linear programming problem. Each of
these regions or phases corresponds to a particular mode of operation
of the metabolic network with qualitatively similar flux patterns within
the phases. In the PhPP analysis there is typically a particular line
separating two phases (line of optimality) that corresponds to the
maximum biomass yield at all parameter values. As described below,
the PhPP approach has proven to be useful for analyzing experimental
data as well as designing experiments [36,37].
Assessing effects of genetic changes . Growth phenotyping studies of
gene deletion strains are a prominent source of data for metabolic
reconstruction validation and expansion, and hence the ability to
predict phenotypes of genetically modified strains is crucial to the
utilization of genome-scale models. The standard FBA approach can be
applied directly to predicting the consequences of gene deletions by
setting the maximum and minimum fluxes through the reactions cat-
alyzed by the deleted gene to zero and repeating the FBA calculation.
However, it is unclear whether the assumption of optimality should
hold for genetically modified strains, and variants of the basic FBA
approach have been proposed to predict deletion phenotypes. In the
minimization of metabolic adjustment (MoMA) approach it is assumed
that, instead of maximizing biomass production, the gene deletion
strain attempts to minimize the adjustment of cellular fluxes from the
wild-type flux distribution [38]. This results in an optimization prob-
lem where the linear objective in problem (5) is replaced by a quadratic
objective representing the Euclidean distance between the wild-type
and deletion strain flux distributions. In the third alternative approach,
regulatory on/off minimization (ROOM), the objective is assumed to
be the minimization of on/off changes of reaction fluxes between
wild-type and genetically modified strains [39].
Evaluating cellular objectives . The objective function used in FBA calcu-
lations is an approximation for the true cellular objective. It is likely
that biological systems are simultaneously trying to satisfy multiple
objectives, and that the objective coefficients c used in FBA do not
represent all these condition-dependent objectives. However, we can
investigate how good a particular objective function is by asking what
objective functions could give rise to an experimentally measured flux
distribution. This problem is essentially the inverse of the FBA problem
(5), since we are now given a flux distribution v and try to find an