Masters candidate in Biostatistics, Jiawei Liu, will present:

“A Novel Penalized Approach for Integrative Modeling of Multiple GWAS Summary Data”

Plan B Adviser: Baolin Wu

Abstract: Given the widespread pleiotropy of genetic variants associated with multiple correlated traits, and the polygenic nature of most complex traits, it is useful to develop joint modeling approach that can aggregate information across multiple traits and variants to improve our power to detect novel genetic variants that are associated with various complex traits. In this paper, we propose a class of multivariate penalized likelihood approach to simultaneously modeling multiple traits and a composite likelihood based empirical Bayes modeling approach to pooling information across genome-wide variants to detect novel variant-trait associations using only the publicly available summary statistics from the genome-wide association studies (GWAS). We derive analytical and numerical algorithms to efficiently solve our proposed genome-wide empirical Bayes penalized estimations. We conduct extensive simulation studies to check that the proposed methods properly control the false positive rates and further illustrate their utility with application to the GWAS summary data for high-density lipoprotein cholesterol (HDL), low-density lipoprotein cholesterol (LDL), and Triglycerides (TG), from the Global Lipids Consortium (GLC) based on around 100,000 European individuals. Our methods identified many novel genes that are missed by the single trait-single variant based GWAS association test. Interestingly, many of them are confirmed to be genome-wide significant in a follow-up GLC study based on around 200,000 European samples. In addition, some of the identified novel genes are worth further study.