Novel Estimation of Genetic Relationship Matrix (GRM) and Comparisons of Different GRMs in Mega- and Meta- GWAS, in the NHLBI TOPMed Program

This dissertation is a collection of statistical tools developed to create genetic relationship matrices for genome-wide association studies, and their application to epidemiological studies of sleep duration in the NHLBI TOPMed program. In chapter one, we proposed a Bayesian method to estimate inbreeding and kinship coefficients, thereby creating a novel Bayesian genetic relationship matrix. In chapter two, we conducted a genome-wide association analysis of sleep duration in 34,840 people representing multiple ancestries in the NHLBI TOPMed consortium. Among the 76M variants and 57,439 gene annotations tested, five variants were found to be significantly associated with sleep duration. In chapter three, we systematically compared the performance of different genetic relationship matrices in linear mixed models for genome-wide association studies, in meta- and mega- analyses. The results supported the Allele-sharing genetic relationship matrix and argued that mega-analyses were superior to meta-analyses.