Epistasis in yeast metabolism relative to metabolic flux phenotypes

Epistasis between two genes indicates a mutual dependency in their impact on phenotype. Previous works have systematically identified epistatic interactions by comparing the phenotypic effects of pairs of single gene deletions and the corresponding double deletions. These studies have shown that epistatic interactions are indicative of functional relationships between genes and can be used to group genes into sets of interacting functional modules. While the sets of interactions identified are likely to be dependent on the observed phenotype, previous genome-scale studies of epistasis have largely been limited to monitoring individual phenotypes, due to the combinatorial complexity inherent in measuring multiple phenotypes across multiple genetic backgrounds. In this dissertation I utilize a genome-scale stoichiometric model of yeast metabolism and the steady-state modeling approach of flux balance analysis, to identify for the first time all epistatic interactions in a biological system, with respect to all observable phenotypes. In particular, I first simulate all 672 single and 225,456 double gene deletions and then identify all epistatic interactions with respect to ∼300 predicted reaction rates as phenotypes.;To justify the use of this modeling approach, I first evaluate available yeast models based on their capacity to predict the metabolic response to gene deletion perturbations. In part one, I assess the ability of three available yeast models to predict the viability of single gene deletion mutants under different environmental conditions. In part two, I evaluate the ability of the best performing yeast model to predict flux changes through central carbon metabolism occurring in response to different gene deletions. Finally, in part three, I use the validated modeling framework to compute a multi-phenotype map of epistasis in yeast metabolism. Analysis of this multi-phenotype map reveals that monitoring different phenotypes yields different sets of epistatic interactions and furthermore, that these different phenotypes can reveal unique aspects of the functional relationships between genes. In addition, I present evidence that patterns of epistatic interactions relative to metabolic flux phenotypes may have constrained the evolution of metabolic enzyme coding genes. This evolutionary signature suggests that the functional dependencies revealed by epistatic interactions place constraints on genome evolution.