Investigating genetic architecture of complex traits through analysis of rare variants and diverse genomes

Abstract

Understanding the relationship between genotype and phenotype is a fundamental goal of genetics, and it has important implications for the etiology of individual differences in behaviors and other complex traits. The current work aims to expand understanding of the genetic etiology of complex human behaviors, such as smoking, through the investigation of under-studied types of genetic variants and under-represented ancestries. In Study 1, whole genome sequencing data from individuals of European ancestry were used to investigate the contribution of rare genetic variants to the heritability of smoking behaviors. Rare genetic variants, especially those present in less than 0.1% of the population, were found to explain non-negligible portions, sometimes comparable to that of common variants, of the phenotypic variance of smoking behaviors. Study 2 developed a novel GWAS trans-ancestry meta-analysis method and performed a large-scale genetic association study of tobacco and alcohol use in diverse ancestries. The study demonstrated that genetic diversity, along with increased sample size, improves locus discovery power and fine-mapping resolution. For most variants, the estimated effect sizes showed limited heterogeneity across ancestries, with several exceptions in the ADH1B and CACNA1B genes. Extending study 2, Study 3 explores how to use diverse genetic data for the estimation of SNP-based heritability and cross-ancestry genetic correlation. We formulate a flexible model to delineate phenotype-genotype relationships in ancestrally heterogeneous samples and then investigate the behaviors of different analytic approaches through simulation and real data analysis. We found that behaviors of these approaches are influenced by cross-ancestry genetic architectures such as the extent to which causal allele frequencies and per-allele size vary across ancestries. We provide guidance in utilizing diverse genetic data in estimating parameters of cross-ancestry genetic architectures and propose ways to define and estimate SNP-based heritability in the presence of effect size heterogeneity along discrete and continuous ancestry measures.