Metabolism is the sum of the chemical interactions occurring inside a cell that process nutrients into cellular constituents and energy. Research in the recent decades has enabled the characterization of metabolic reaction networks for multiple organisms. Genome-scale models (GEMs) have been utilized to mathematically describe and analyze these networks. GEMs are stoichiometric models that can serve as basis for studying metabolism and identifying metabolic states. However, stoichiometric models of metabolism do not account for the dynamics of the system. An understanding of cellular dynamics and regulation is essential for metabolic engineering of cells that produce certain metabolites of interest. Kinetic models can provide invaluable knowledge about system dynamics of cellular metabolism as metabolic control analysis (MCA) can offer information about the systemâ s response to perturbations. Nonetheless, the construction of kinetic models is a challenging endeavor as several hurdles have to be overcome. Kinetic models of metabolism are subject to two general issues: diversity of network topologies and underlying uncertainty. Kinetic models are often built on an ad hoc basis without clear explanation or justification about their network contents, as there is no systematic protocol for their construction. Furthermore, kinetic modeling is subject to uncertainty from various sources. Multiple steady states can characterize an observed physiology. Additionally, the kinetic mechanisms describing a system and the values of their kinetic parameters are often not known.

In this thesis, we tackle several issues that hinder the formulation of kinetic models. Firstly, we apply model reduction algorithms to systematically reduce GEMs to construct study-specific kinetic models of different degrees of complexity. We demonstrate that the MCA outputs for these kinetic models are mostly independent of the complexity level because we preserve model equivalency. Secondly, as published kinetic models to date are constructed around a steady state of metabolism, we analyzed the impact of alternative steady states on metabolic engineering strategies. We proposed a systematic workflow for deriving conclusions that take into account the alternative feasible steady states of a physiology. We concluded that MCA outputs are more sensitive to the metabolite concentrations than to the metabolic fluxes of the system. Hence, it is essential to consider alternative metabolic states for a given physiology. Thirdly, since multiple kinetic models can describe a physiology due to the given uncertainties, it is important to consider them in populations. In order to derive conclusions from populations of kinetic models, it is essential to quantify with what certainty we make such predictions. We tested different statistical methods for assigning confidence levels to our conclusions and make recommendations applicable to any kinetic models. Finally, we implemented a sensitivity analysis approach that can elucidate which input parameters of a kinetic model contribute the most to the uncertainty in MCA outputs. The approach appears to predict correctly the sources of uncertainty and can be applied to any large-scale kinetic models. Overall, the work from this thesis contributes towards establishing a systematic workflow for building more consistent and comprehensible kinetic models for observed physiologies under uncertainty.