UC Berkeley

Adaptation is a fundamental process in evolution, which leads populations to better survive and reproduce in changing environments. A key insight of population genetics has been that present-day genetic variation is affected by past, and even ongoing, adaptations. Recent expansion of DNA sequencing has afforded us access to genetic variation from now up to nearly millions of individuals. In this dissertation, I develop modeling and inference for DNA sequence variation in order to identify the genetic bases of adaptations, with a focus on evolution of complex traits in humans. First, I review population genetics approaches for detecting natural selection, and argue that these approaches have been hamstrung by the intractability of the so-called ‘full likelihood’ of selection (Chapter 1); I then develop a method to tractably compute this likelihood via importance sampling of the ancestral recombination graph (ARG), enabling us to find targets of selection too subtle to detect with previous methods (Chapter 2); I extend this likelihood method to jointly model DNA sequence variation and complex trait variation (via summary statistics from genome-wide association study, aka GWAS) to quantify the amount of selection acting on a complex trait, and to account for pleiotropy/correlated response in these estimates (Chapter 3); Finally, I present a method to detect polygenic adaptations in the presence of population structure, which explicit accounts for uncorrected stratification and other sources of error in the GWAS via an approach similar to LD score regression (Chapter 4).