(482b) Optimizing Ensemble Modeling Framework to Generate Kinetic Models of Metabolism

Enabling kinetic and regulatory modeling of cellular metabolism is a major challenge in metabolic engineering and systems biology. Constraint-based stoichiometric modeling greatly aids in characterizing and improving strain designs but without kinetic information it is difficult to identify rate limiting steps and interrogate regulatory behavior. Unfortunately, in vitro derived kinetic parameters for enzymes do not necessarily reflect true in vivo behavior and are often determined under varying experimental conditions. Consequently, a single kinetic model combining these in vitro derived parameters is often unable to resolve experimentally observed in vivo data. The ensemble modeling (EM) framework was previously developed to address these hurdles by sampling kinetic parameters for the entire metabolic network simultaneously and screening them against a single experimental dataset.[1] Furthermore, the EM method constrains the large kinetic parameter sample space using readily available thermodynamic, stoichiometric, and steady state flux data. However, despite its numerous advantages, EM becomes computationally limiting with increasing network size and complexity. As we increase the availability of larger stoichiometric metabolic models for various organisms, it is imperative to improve our ability to generate larger kinetic models of these systems as well.

In this work we build on the previous developments in EM by optimizing parameter screening techniques and introducing methods to reduce structural model complexity. We have investigated the trade-offs between model fitness and computation time associated with additional parameter sample space constraints, alternative model screening methods, and reduced kinetic rate law forms. We have also implemented a conservation analysis step to eliminate linear dependency in our models and reduce the stiffness and screening time of our kinetic parameter sets. Most importantly, our work toward optimizing the existing EM method can directly plug into and benefit concurrent EM efforts to develop genome-scale kinetic models across our field.[2] Specifically, through reducing computation time and optimizing parameter sampling and model screening, we are better equipped to more extensively explore the larger parameter sample spaces associated with larger metabolic networks. Lastly, we hope our combined findings will serve as a useful starting point for future EM users to select the best sampling and screening method options to use in their own, unique applications.

[1] Tran, L. M., Rizk, M. L., & Liao, J. C. (2008). Ensemble modeling of metabolic networks. Biophysical journal, 95(12), 5606-5617.

[2] Khodayari, A., Zomorrodi, A. R., Liao, J. C., & Maranas, C. D. (2014). A kinetic model of Escherichia coli core metabolism satisfying multiple sets of mutant flux data. Metabolic engineering, 25, 50-62.