Thermodynamic Framework for Mutant Phenotype Prediction

Oyetunde, T., Washington University in St. Louis
Tang, Y. J., Washington University
Accurate prediction of the metabolic flux redistribution within mutant strains is vital for directing rational strain design. The discrepancies between experimentally determined and computationally predicted flux distributions imply that the predictive techniques do not accurately capture the mechanisms of cellular regulation.

Previous mutant prediction algorithms have essentially two components: (1) a metric to characterize the cell’s desired metabolic state (for example flux or gene expression profiles) and (2) a metric to describe the distance from the desired state (for example, Euclidean distance). The mutant flux profile is then computed as the closest possible to the wild type state subject to the constraints of genetic or environmental perturbations. Different algorithms have reported significant gains in accuracy by tweaking the definitions of the two metrics.

To further improve the fidelity of knockout predictions and subsequent computational strain design, we developed a thermodynamics-based method, RElative MEtabolite Patterns (REMEP). REMEP hypothesizes that the optimum metabolic state is reflected in the energetic requirements to sustain flux through each metabolite node, and thus cell fluxomes adapt to perturbations from a reference state by preserving relative pattern of metabolite energy flows. REMEP performs better than comparable algorithms across different experimental datasets for E. coli and S. cerevisiae (in terms of lower root mean square errors and higher Pearson’s correlation coefficients). These improvements support the REMEP assumption that cellular mechanisms of response to genetic and environmental perturbations leaves signatures that can be inferred from thermodynamics-derived metabolite patterns. The findings provide a new paradigm for genotype to phenotype mapping and insights into microbial flux network plasticity.