Abstract
Detailed dynamic kinetic models at the network reaction level are traditionally constructed using
mechanistic enzymatic rate equations and a large number of kinetic parameters have to be determined under nonphysiological
conditions in vitro. However, the validity of these parameters under in vivo conditions is doubtful
and the rates equations are usually highly complex. Therefore, one of the major obstacles in building accurate
kinetic models is the lack of detailed knowledge of the rate laws that describe the reaction mechanism and the
absence of their associated parameters. There is an urgent need for alternative modelling approaches to fill this
gap. In this study, we analyze four alternative hybrid modeling strategies to the reference large scale mechanistic
E. coli central carbon metabolic network model based on the Michaelis-Menten equation only for the bimolecular
reactions and the other reactions with different formats of approximative rate kinetics (Generalized Mass-Action,
convenience equation, lin-log and power-law). These rate equations help to reduce the number of parameters that
have to be estimated. The kinetic parameters optimization was performed through the combination of a global
search evolutionary programming method followed by a local optimization method (Hooke and Jeeves) to refine
the fitting. Predictions and stability analyses to test the viability of the alternative models were also performed.
The good dynamic behaviour and powerful predictive power obtained by the mixed modeling composed on
Michaelis-Menten kinetics and the approximate lin-log kinetics indicate that this as a suitable approach to
complex large scale models where the exact rate laws are unknown.
Keywords:
Large-scale E. coli metabolic network
Dynamic modeling
Approximative enzyme kinetics
Parameter estimation
Publication Type: Abstracts in Conference Proceedings