Leveraging gene expression and local ancestry to investigate regulatory epistasis in humans

Abstract

Epistasis is a phenomenon wherein the effect of a genetic variant on a phenotype is dependent on the genomic context. Better understanding epistastic relationships between variants, often termed interactions, can shed light on novel genomic loci associated with complex disease, which may improve our understanding of the underlying biological mechanisms. Additionally, capturing epistastic effects in models of disease risk may help improve predictions of at-risk populations, or the prediction of a variant’s deleteriousness in precision medicine initiatives. However, the study of epistasis faces unique methodological challenges, and consequently, evidence for regulatory epistasis remains elusive in humans. In this work, I address two major challenges within the field of regulatory epistasis: the development of statistical best practices, and the investigation of epistasis within haplotypes. In Chapter 2, I illustrate that traditional quality control procedures are insufficient to correct for confounding in studies of epistasis, and develop a set of additional guidelines for future studies. Once these were applied, I found little evidence for epistasis between common, unlinked variants influencing gene expression levels. In Chapter 3, I leverage unique properties of admixed populations to investigate epistasis within ancestral haplotypes disrupted by ancestry-specific recombination events. I find several examples of epistasis with plausible biological support, which serve as a proof of principle for the utility of this approach. Overall, these findings indicate that regulatory epistasis likely has small effects, occurs within haplotypes, or occurs between distant genomic regions; we recommend future studies of epistasis focus on these possibilities.