Programming Biobank Scale Identity-by-Descent Analysis

Abstract

Accurate characterization of familial relationships and subsequent genetic association detection remain fundamental challenges in large-scale genomic studies, due to both underutilization of available methods and scale and diversity of modern input data. Pedigree reconstruction and assessments of relatedness, vital precursors to investigations of disease etiology, become complicated as extant tools face high computational demands, reduced accuracy in admixed populations and increased sample missingness. New tools for disease gene discovery are also needed, as genome-wide association studies (GWAS) are underpowered to detect genetic variants with high allelic heterogeneity and traditional family-based approaches such as linkage are underpowered in the context of high genetic heterogeneity, rendering a class of causal variants undetectable by popular approaches. To overcome these limitations, my dissertation work utilizes patterns of identity-by-descent (IBD) for new and improved genomic discovery in large, diverse populations. First, I present Combined Pedigree-Aware Distant Relatedness Estimation (COMPADRE), a method integrating multiple IBD estimation approaches to improve accuracy and efficiency of pedigree reconstruction, especially in complex pedigree structures. By accurately reconstructing familial relationships from genetic data alone, COMPADRE can power robust inheritance pattern analysis and improve heritability estimation. Second, I introduce IBDMap, a novel tool for detecting genomic regions enriched with IBD segments among affected individuals. Third, I demonstrate that IBDMap successfully identifies loci harboring rare disease-associated variants that traditional methods have missed and expands the possibility of disease gene discovery in novel biobank settings. Together, these tools expand our ability to uncover novel genetic associations, improve diagnostic precision, and ultimately advance precision medicine.