Data Simulation to Optimize the GWAS Framework in Diverse Populations

Whole-genome or genome-wide association studies have become a fundamental part of modern genetic studies and methods for dissecting the genetic architecture of common traits based on common polymorphisms in random populations. It is hoped that there will be many potential uses of these identified variants, including a better understanding of the pathogenesis of traits, the discovery of biomarkers and protein targets, and the clinical prediction of drug treatments for global health. Questions have been raised on whether associations that are largely discovered in populations of European descent are replicable in diverse populations, can inform medical decision-making globally, and how efficiently current GWAS tools perform in populations of high genetic diversity, multi-wave genetic admixture, and low linkage disequilibrium (LD), such as African populations. In this study, we employ genomic data simulation to mimic structured African, European, and multi-way admixed populations to evaluate the replicability of association signals from current state-of-the-art GWAS tools in these populations. We then leverage the results to discuss an optimized framework for the analysis of GWAS data in diverse populations and outline the implications, challenges, and opportunities these studies present for populations of non-European descent.