Bayesian Inference of Metabolic Kinetics from Genome-Scale Multiomics Data

Optimizing the metabolism of microorganisms for maximum yields and titers is a critical step in improving bioprocess economics. With the growing availability of transcriptomic, proteomic, metabolomic, and fluxomic analysis techniques, characterization of engineered strains has become increasingly detailed. However, utilizing multiomics data to make informed decisions about future strain improvements remains a major challenge in modern bioengineering. Parameterizing kinetic models from indirect, in vivo data is typically infeasible at the genome-scale. There is therefore a need for mechanistic modeling frameworks that can consume the large amount of data generated through multiomics experiments to yield actionable insight for strain engineers.

Metabolic ensemble modeling has emerged in recent years as an effective tool for estimating confidence intervals in kinetic metabolic models from observable steady-state flux and concentration measurements. The method operates by sampling feasible kinetic parameters and filtering to sets of parameters that match experimental observations. However, these methods rely on an explicit numerical integration of large kinetic models, and therefore scale poorly both in the size of the network that can be modeled, as well as the amount of data that can be used to constrain the parameters.

We demonstrate the coupling of linear-logarithmic kinetics to a Bayesian inference framework. The resulting method therefore represents a scalable, flexible method for modeling omics data. Specifically, linear-logarithmic kinetics enable steady-state fluxes to be predicted linearly from kinetic parameters, removing the computational burden associated with solving for steady-state flux. Additionally, since linear solutions permit the easy determination of likelihood gradients, advanced Bayesian inference techniques can be applied to reliably estimate posterior parameter distributions even for high-dimensional models. We demonstrate the method through a number of case studies, from simple in vitro reaction systems to complex, medium-scale metabolic models with hundreds of metabolites and reactions parameterized with modern multi-omic data.