Multiomics Integration for Prediction of Complex Metabolic Phenotypes

A key challenge in solving genotype to phenotype relationship is to predict a cell’s metabolome. Here, we quantified the proteomes in 97 non-essential kinase knock-out strains from Saccharomyces cerevisiae and associated these to their metabolomes. We observed that kinase knockouts affect proteomes broadly but in a distinct manner; and that these are quantitatively dominated by expression changes in metabolic enzymes. Analysing these data in the context of kinetic modeling demonstrated that enzyme abundance affects metabolite concentrations through the redistribution of flux control, resulting in a many-to-many relationship between enzyme abundances and the metabolite concentrations. Machine learning enabled mapping these relationships, the prediction of experimentally measured metabolite concentrations, as well as to identify candidate genes important for the regulation of metabolism. Overall, our study suggests that hierarchical metabolism regulation acts predominantly through adjustment of broad expression patterns rather than over individual rate-limiting enzymes, and may account for more than half of metabolism regulation.