Integrating Transcriptomic and Metabolomic Data with Thermodynamically Consistent Metabolic Models

Constraint based models are a powerful mathematical framework for integrating experimental data from different sources. In this study we demonstrate the added value in combining various levels of data. The first is metabolomics and thermodynamics via Thermodynamics-Based Flux Analysis (TFA) constraints. TFA constraints guarantee the thermodynamic consistency between measured metabolomics levels and flux directionalities. We further introduce transcriptomics-based constraints (derived from IMAT), that allow us to integrate transcriptomics together with TFA constraints for the first time. We apply the methodology in an E. coli genome scale model using available metabolomics and transcriptomics data.

Despite integrating several levels of data, the problem is underdetermined and there was a multitude of alternative solutions that can explain the observed\integrated data. Oftentimes results are only reported for a single solution that is chosen arbitrarily, and alternative possible solutions are ignored. In the studies here we also calculated alternate optimal solutions and we showed that alternative solutions can differ drastically different from one another and they should always be taken into account in order to have unbiased predictions.

We also found that transcriptomics-based predictions differ with and without thermodynamics. Without thermodynamics-based constraints, we found that the transcriptomics-based constraints predict thermodynamically infeasible fluxes at optimality. When we forced consistency with both transcriptomics and metabolomics data sets simultaaneously, we produced more precise and accurate predictions of metabolic fluxes, according to C13 measurements. These results further emphasize that all models should be thermodynamically consistent.