Development and application of polygenic risk scores for coronary heart disease.

Abstract

This dissertation explores two approaches to enhance the predictive power of polygenic risk scores (PGS) for coronary heart disease (CHD) and other cardiometabolic traits, with a particular focus on improving their applicability across diverse populations. We present two strategies for PGS improvement: (1) leveraging summary statistics from genetically correlated traits and publicly available PGS across the phenome to create more predictive PGS for CHD (multi-PGS), and (2) integrating functional genomic annotations, such as cell type-specific snATAC-seq peaks and fine-mapped tissue-specific eQTLs, with GWAS data to enhance PGS performance for cardiometabolic traits. We demonstrate that multi-PGS for CHD have increased predictive power. Additionally, the integration of functional genomic annotations with GWAS data resulted in an average 13% improvement in PGS performance for cardiometabolic traits, with greater improvements observed in individuals of African ancestry. The dissertation also explores two practical applications of PGS for CHD. First, using an angiography dataset from Mayo Clinic patients, we show that PGS for CHD associate continuously with the extent of atherosclerosis, suggesting its utility in predicting disease severity beyond binary outcomes. Second, we investigate the interplay between PGS for CHD and social determinants of health (SDOH) across the United States, revealing that SDOH associate with CHD independently of PGS, highlighting the importance of considering both genetic and environmental factors for equitable risk assessment. This research contributes to the ongoing effort to improve the accuracy and equity of genetic risk prediction in cardiovascular medicine. It underscores the need for diverse cohorts in PGS development, the potential of functional genomic data in enhancing predictive models, and the critical importance of integrating SDOH alongside genetic factors for comprehensive and equitable risk assessment.