UC San Diego

Most genetic studies have historically centered on single-nucleotide polymorphisms (SNPs) in individuals of European ancestry, overlooking both complex genomic variants and global diversity. This thesis focuses on improving our understanding and analysis of tandem repeats (TRs)—a major, yet underexplored, source of genomic variation—through a multi-ancestry lens. First, we developed EnsembleTR, a genotyping framework that harmonizes TR calls across methods, and used it to construct a reference panel of 1.7 million TRs from 3,550 individuals in the 1000 Genomes Project and H3Africa cohorts. This panel enabled various analyses, such as identifying ancestry-specific repeat expansions, expression quantitative trait loci (eQTL) mapping, and investigating sequence determinants of TR stability. To further improve our understanding of TRs, we built LongTR, a new tool for accurate TR quantification from both PacBio and Oxford Nanopore data, demonstrating superior accuracy and scalability over existing tools. Finally, we investigated two novel approaches for constructing multi-ancestry polygenic risk scores (PRS). First, we developed imPRS, that jointly models posterior effect sizes across populations and introduces a correlation parameter to enforce similar effect sizes for the shared variants among population. Second, we explored building the PRS model without reliance on ancestry labels in genome-wide association studies (GWAS), linkage disequilibrium (LD) estimation, and PRS calculation, by pooling all samples from every ancestry into one single cohort. Together, these contributions advance the inclusion of structurally complex variants and diverse populations in genetic analyses, with implications for improved variant discovery and equitable risk prediction.