UC San Diego

Evolutionary dynamics depend on how segregating alleles affect organismal fitness. The ability to predict such dynamics would have far reaching applications from human health to agriculture and biotechnology. A major obstacle to understanding them is that the fitness benefit provided by an allele in one genetic background in one environment will often change, and may even turn into a fitness loss, if this allele lands in a different genetic background (G×G interactions, or, epistasis), a different environment (G×E interactions, or, pleiotropy) or both (G×G×E interactions). Novel technologies have enabled increasingly high-throughputmeasurements of mutational effects, which allow us to empirically understand some G×G, G×E and G×G×E interactions. However, using these measurements directly to predict evolutionary dynamics is challenging for several reasons. In particular, the set of available beneficial mutations changes as alleles fix in an evolving population, and the environment may change over the course of adaptation. As such, the number of relevant interactions to measure grows exponentially with evolutionary time.

In this dissertation, I develop a unified framework with which to understand the longterm evolutionary consequences of pleiotropy and epistasis using a combination of simulations, mathematical modeling, and empirical data analysis. In Chapter 1, I develop a novel framework called the “Joint Distribution of Fitness Effects” (or, JDFE), which characterizes how mutations in one environment impact fitness in others. I demonstrate how the JDFE determines evolutionary dynamics with an application to the evolution of muti-antibiotic resistance. In Chapter 2, I analyze a large dataset of mutational effects in diverse strains of yeast Saccharomyces cerevisiae to uncover simple predictive patterns in epistasis across environments. In Chapter 3, I extend the empirical cross-environmental epistasis model from Chapter 2 to predict how JDFEs change with evolutionary time. I explore the implications of this unified model to the predictability of long-term evolutionary dynamics in variable environments.