A more powerful quasi-likelihood score test for detecting genetic association with multivariate phenotypes in related samples

Abstract

Pleiotropy is a commonly observed phenomenon in human genetics where a single gene influences multiple, and sometimes seemingly unrelated traits. Recently there has been significant interest in the identification of genetic variants that are associated with multiple phenotypes since identifying pleiotropic effects can lead to a better understanding of the underpinnings of complex traits. Genome-wide association studies often collect data on a variety of phenotypes, and a number of methods have been proposed for the joint analysis of multiple phenotypes in unrelated samples. Many genetic studies, however, include related individuals. In this thesis, we consider the problem of genetic association testing with multivariate phenotypes in samples with relatedness. We propose the multivariate phenotype quasi-likelihood (MPQ) score test for association mapping in related samples. Some of the features of the MPQ are: (1) it is applicable to completely general combinations of family and population-based samples, (2) it allows for the analysis of general quantitative traits and can accommodate both binary and continuous outcomes, (3) it can incorporate information on covariates in the analysis, and (4) it is computationally feasible for large-scale GWAS allowing for arbitrary relatedness among sample individuals. In simulation studies with unrelated and related samples, we demonstrate that the MPQ represents an overall, and in many cases, substantial, improvement, over existing multivariate methods, in terms of type-1 error rate and power, for a variety of causal models and multivariate trait correlation structures. Finally, we apply the MPQ test to a GWAS of 3,548 Hispanic American postmenopausal women from the Women’s Health Initiative SNP Health Association Resource to identify genetic variants associated with pleiotropic effects on serum C-reactive peptide (CRP) and white blood cell counts (WBC), two inflammation-related phenotypes. The MPQ test identifies previously reported variants for CRP and WBC as well as novel variants that are genome-wide significant.