Genetic Prediction of Complex Traits Across Diverse Populations

Background: Discoveries made through genome-wide association studies have revolutionized the field of human genetics by uncovering disease mechanisms and enabling precision medicine. Although there has been tremendous effort to obtain cohort sizes on the order of hundreds of thousands of individuals, there is an immense underrepresentation of non-European ancestries, which has potential to contribute to health inequity. Here, we present works on (i) an in-depth simulation to identify optimal approaches for achieving equitable accuracy for polygenic risk scores (PRS) in diverse populations and (ii) an application of trans-ancestry discovery of germline associations in the complex phenotype of multiple primary tumors. Methods: (i) Through our simulation framework, we implement many strategies for building PRS, including a local-ancestry specific approach, and measure accuracy in admixed and African ancestry individuals. (ii) We also conducted a whole-exome sequencing study of two large, multi-ancestry populations consisting of 6,429 multiple cancer cases, 29,091 single cancer cases, and 165,853 cancer-free controls. We employed single-variant and gene-based tests to characterize the genetic susceptibility to multiple primary tumors in comparison to individuals with one and, separately, no cancers through the investigation of rare and common variation. Results and Conclusions: (i) Variants discovered in African ancestry populations have greater potential to achieve unbiased PRS prediction across populations. Studies should prioritize the inclusion of diverse participants in GWAS, and care must be taken with the interpretation of currently available risk scores. (ii) Our applied trans-ancestry analysis of multiple primary tumors identifies rare loss-of-function variants and gene-level associations with cross-cancer pleiotropy and potential for prioritizing cancer survivors at high risk for developing subsequent tumors