Developing a seamless pipeline from 13C labeling data to kinetic models with a genome-wide coverage

Kinetic models of metabolic networks offer the potential of truly predictive models of metabolism. The mechanistic characterization of enzyme catalyzed reactions allows for accurate prediction of perturbations in metabolite concentrations and reaction fluxes in response to genetic and environmental perturbation that are beyond the scope of stoichiometric models. Despite their potential, the application of kinetic models to microbial strain design and metabolic engineering has been limited by (i)the paucity of fluxomic datasets that span the entire metabolism needed for parametrization (ii)the computational expense associated with kinetic model parameterization.

In this talk, we highlight a two-step pipeline that promises to alleviate some of the inherent computational challenges. It requires as an input a curated genome-scale-metabolic(GSM) model and 13C-labeling distributions for intracellular metabolites under multiple genetic and environmental perturbations. Additional data such as biomass yield, metabolite concentrations, enzyme kinetics and/or fermentation data can be integrated whenever available. The key concept here is to use the same GSM metabolic map for performing both metabolic flux elucidation and kinetic model parameterization. This eliminates any errors due to ad-hoc model aggregation/simplification. The first-step involves the application of improved algorithms for performing both stationary and instationary MFA using atom mapping model with genome-wide coverage. The second-step uses as input the elucidated metabolic flux ranges to estimate kinetic parameters that agree with all fluxomic datasets using the newly developed K-FIT-parameterization algorithm. The K-FIT-algorithm relies on a combination of nested decomposition schemes and iterative solution techniques to evaluate steady-state fluxes in response to genetic perturbations. It achieve orders of magnitude CPU improvements compared to a genetic-algorithm(GA)-based procedure. The approach has been tested for a core model of E.coli containing 109-reactions and 61-metabolites, a large-scale E.coli model with 319-reactions and 276-metabolites as well as models for clostridia. Using this workflow, 13C-data can be seamlessly translated into parameterized kinetic models.