Nonparametric Bayesian Variable Selection for Gene-Environment Interactions in Genome-Wide Association Studies

• Activity Number: 74 - Invited E-Poster Session I
• Type: Invited
• Date/Time: Sunday, August 7, 2022 : 8:30 PM to 9:25 PM
• Sponsor: Biometrics Section
• Author(s): Xiaoxi Li and cen wu*
• Companies: Kansas State University and Kansas State University
• Keywords: Genom-Wide Association Study; Markov Chain Monte Carlo; nonparametric Bayes; ultra-highdimeisonal data; variable selection

Abstract

In the Genome-Wide Association Studies (GWAS), a major objective is to identify susceptible variants underlying disease traits. Gene-environment interactions have been historically acknowledged as a major player for better understanding of the genetic architecture of complex diseases including cancers. Nevertheless, detection of important gene-environment (GxE) interactions is especially challenging given the ultra-high dimensionality of GWAS. In this study, we propose a nonparametric Bayesian variable selection method for GxE interactions in GWAS. Our method enjoys a marginal nature and is scalable to GWAS even though it has been developed based on Markov Chain Monte Carlo (MCMC). In addition, the proposed method can handle different forms of GxE interactions through nonparametric modelling. The performance of the proposed approach is demonstrated through simulation studies and a case study of GWAS.