Dissertations & Theses (Open Access)

Degree Name

Doctor of Philosophy (PhD)


Han Chen


Genetic association tests have enabled people to identify susceptible loci and broadened our understanding of complex diseases. However, the GWAS (Genome-wide Association Studies) results are probably confounded by population stratification thus leading to potential false-positive findings. This problem is pronounced particularly in admixed populations, a group of populations with multiple ancestries whose local genetic ancestry may drastically vary at local genomic positions (local ancestry) compared to the overall genetic ancestral composition (global ancestry). It is insufficient to only account for global population structure in admixed populations. Methods have been developed to account for local population stratification but followed by subsequent issues, such as heavy computational workload, to refit null models wherever local ancestry changes. In this dissertation, we aimed to develop a local-ancestry-aware GWAS model that comprehensively adjusts for population stratification effects via local ancestry in admixed populations. We first proposed a new continuous measure to delineate local ancestry patterns that is able to extract and refine ancestral characteristics in relatively low dimensions. Our measure was accurate and robust against local ancestry misclassifications. We conducted large-scale simulation tests with both accurately inferred and misclassified local ancestry estimates to evaluate the performance of our newly proposed continuous local ancestry measure. We next incorporated the new continuous measure into a local-ancestry-aware GWAS model to efficiently adjust for confounding by local ancestry. We demonstrated its superior performance over conventional GWAS models that do not account for local ancestry by conducting large-scale type-I error simulations. We also applied the newly developed model to a local-ancestry-aware GWAS on physician-diagnosed asthma in the Hispanic Community Health Study / Study of Latinos (HCHS/SOL). We successfully identified a strongly associated locus on chr6 that was not previously discovered by conventional GWAS models in Hispanic/Latino Americans. Our newly proposed continuous local ancestry measure and local-ancestry-aware GWAS models effectively control for confounding of population stratification, and we expect our computationally efficient methods to facilitate genetic association studies on complex diseases and their quantitative risk factors in admixed populations.

Available for download on Monday, October 28, 2024