The effects of model complexity and size on metabolic flux distribution and control. Case study in E. coli

Abstract

Significant efforts have been made in building large-scale kinetic models of cellular metabolism in the past two decades. However, most kinetic models published to date, remain focused around central carbon pathways or are built around ad hoc reduced models without clear justification on their derivation and usage. Systematic algorithms exist for reducing genome-scale metabolic reconstructions to build thermodynamically feasible and consistently reduced stoichiometric models. It has not been studied previously how network complexity affects the Metabolic Sensitivity Coefficients (MSCs) of large-scale kinetic models build around consistently reduced models. We reduced the iJO1366 Escherichia Coli genome-scale metabolic reconstruction (GEM) systematically to build three stoichiometric models of variable size. Since the reduced models are expansions around the core subsystems for which the reduction was performed, the models are modular. We propose a method for scaling up the flux profile and the concentration vector reference steady-states from the smallest model to the larger ones, whilst preserving maximum equivalency. Populations of non-linear kinetic models, preserving similarity in kinetic parameters, were built around the reference steady-states and their MSCs were computed. The analysis of the populations of MSCs for the reduced models evidences that metabolic engineering strategies - independent of network complexity - can be designed using our proposed workflow. These findings suggest that we can successfully construct reduced kinetic models from a GEM, without losing information relevant to the scope of the study. Our proposed workflow can serve as an approach for testing the suitability of a model for answering certain study-specific questions. Author Summary Kinetic models of metabolism are very useful tools for metabolic engineering. However, they are generated ad hoc because, to our knowledge, there exists no standardized procedure for constructing kinetic models of metabolism. We sought to investigate systematically the effect of model complexity and size on sensitivity characteristics. Hence, we used the redGEM and the lumpGEM algorithms to build the backbone of three consistently and modularly reduced stoichiometric models from the iJO1366 genome-scale model for aerobically grown E.coli. These three models were of increasing complexity in terms of network topology and served as basis for building populations of kinetic models. We proposed for the first time a way for scaling up steady-states of the metabolic fluxes and the metabolite concentrations from one kinetic model to another and developed a workflow for fixing kinetic parameters between the models in order to preserve equivalency. We performed metabolic control analysis (MCA) around the populations of kinetic models and used their MCA control coefficients as measurable outputs to compare the three models. We demonstrated that we can systematically reduce genome-scale models to construct kinetic models of different complexity levels for a phenotype that, independent of network complexity, lead to mostly consistent MCA-based metabolic engineering conclusions.