Statistical Methods in Admixture Mapping: Mixed Model Based Testing and Genome-wide Significance Thresholds

Abstract

Genetic admixture occurs when two or more previously isolated populations combine to form an admixed population. The study of admixed populations can provide valuable insights into the complex relationship between environmental exposures, genetic background and complex traits. Gene mapping by linkage admixture disequilibrium, or admixture mapping, is a powerful approach for the identification of genetic loci influencing complex traits in ancestrally diverse populations. Admixture mapping leverages genomic heterogeneity among sampled individuals for improved gene discovery, where genetic loci with unusual deviations in local ancestry and that are significantly associated with a trait are identified. Admixture mapping can serve both as a primary method for discovery of novel genetic variants and as a complement to association mapping. In this dissertation, we thoroughly investigate the performance of existing statistical methods used for admixture mapping and we develop new methods that improve upon existing approaches. We also characterize the correlation structure of genetic loci in admixed populations and develop new genome-wide significance thresholds for admixture mapping under a range of models that should be useful for the future studies. Using real genotyping data in a large sample of African Americans, we find evidence of assortative mating, and in simulation studies with simulated phenotypes, we demonstrate that ancestry-related assortative can induce genome-wide inflation of admixture mapping test statistics and false positive associations. We also show how to appropriately adjust for this inflation and protect against spurious admixture associations. Finally, new linear and logistic mixed model methodology is developed for admixture mapping of quantitative and binary traits, respectively, in the presence of relatedness and population structure. We evaluate the performance of these methods through extensive simulation studies. The methods are applied to large-scale genetic studies of African American and Hispanic/Latino populations for genome-wide admixture mapping analyses where novel candidate loci for a variety of biomedical traits are identified.