Large Scale Dynamic Models of Metabolism as a Tool for Analysis and Design of Metabolic Engineering and Synthetic Biology Strategies

Large-scale dynamic models of metabolism remain an overarching ambition of efforts that aim to analyze and explain complex behavior of living entities. However, intrinsic nonlinearity of dynamics of living organisms and structural and quantitative uncertainties at several biological levels hinder the development of such models. Specifically, the knowledge about kinetic mechanisms of the metabolism is still limited and the corresponding parameters are frequently not available. Moreover, identification of the missing parameters would necessitate a prohibitively large number of experimental measurements. We are presenting a systematic methodology for generating a population of large-scale dynamic nonlinear models of metabolism. We start from the reference steady-state flux and we use the ORACLE[1] (Optimization and Risk Analysis of Complex Living Entities) framework to integrate available metabolomics and kinetic data into these dynamic descriptions and we computed the corresponding parameters such as Michaelis constants and maximal velocities. The parameter estimation method proposed here is tailored for metabolic kinetic models and can be used as an alternative to standard parameter estimation methods in the biologically relevant cases when the parameter estimation problem does not have an unique solution. In addition, the obtained models are locally stable around the reference state and consistent with the physico-chemical laws and thermodynamics. We demonstrated the potential and features of the proposed methodology for the case of optimally grown E. coli, where we constructed a population of large-scale nonlinear models that involved 283 metabolites and 409 reactions. For the analyzed physiology we: (i) identified and analyzed multiple steady-states; (ii) characterized basins of attraction around the identified steady-states; and (iii) investigated the responses of E. coli metabolism upon large perturbations such as gene knockouts. We further investigated the potential of the proposed methodology for the design of metabolic engineering and synthetic biology strategies.