Systems Biology and Metabolic Engineering in Bacteria - Systems Biology of Metabolic and Signaling Networks - page 349

dynamics of metabolic networks, simply because quantitative data about enzymatic
reactions are difficult to obtain. Furthermore, such data are usually measured
in vitro and their relevance for the in vivo situation remains questionable. Attempts
have therefore been made to extend the genomic-scale, steady state metabolic
models by adding dynamics in different ways. As mentioned above, FBA, for
example, strongly depends on the objective function and the constraints imposed
on the system. Since these change during a typical growth experiment, making them
time dependent leads to dynamic flux balancing. Meadows and colleagues have
used this approach to successfully model E. coli fermentation in an industrial
bioreactor, using FBA and time-dependent inputs, such as rate-dependent biomass
composition (Meadows et al. 2010 ).
In principle, the data used for steady state network analysis could also be used to
derive kinetic constants, and therefore a dynamic model of the network. Classic
dynamical models are formulated as systems of nonlinear differential equations.
Recovering all parameters of such a complicated system seems overambitious.
However, in many situations, the nonlinear differential equation system is well
approximated by lin-log, power-law, or S-system descriptions (Heijnen 2005 ).
Using high-throughput data (Ishii et al. 2007 ), where substrate and reactant
concentrations were quantified along with corresponding reaction fluxes in a series
of steady state perturbation experiments, Berthoumieux et al. ( 2012 ) explored the
possibility of identifying the parameters of the corresponding lin-log model. They
concluded that even with such an extensive dataset and the linearizing model
simplifications, only four out of 31 reactions, and 37 out of 100 parameters were
identifiable.
Another approach was to take established steady state, genome-scale models and
incorporate available kinetic information (Jamshidi and Palsson 2010 ). Enzymes
are represented explicitly in this formulation. Different steady state measurements
are used to estimate equilibrium constants, and subsequently kinetic constants of
individual enzymatic reactions. These constants are then incorporated in the kinetic
model using bilinear equations. This model faithfully describes system responses to
external perturbations such as changes in the energy or redox state of the cell. The
modeling algorithm makes extensive use of -omics data. Model construction thus
profits from the vast amount of such data that are publicly available and avoids the
“tedious” measurement of individual reaction constants necessary for the classical
modeling approach. However, classical enzyme kinetic models, when available, are
still a more faithful representation of system kinetics. Important biological phe-
nomena, such as regulatory interactions via metabolites or second messengers,
cannot be represented by the intrinsically linear approximation of FBA. Moreover,
collective phenomena such as self-organization or synchronized oscillations are
certainly beyond the realm of MFA or FBA.