Network-Based Bayesian Variable Selection Approach to Genome-Wide Association Studies Data

• Activity Number: 15
• Type: Topic Contributed
• Date/Time: Sunday, July 31, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Peng Wei*+
• Companies: The University of Texas School of Public Health
• Keywords: Bayesain variable selection ; gene networks ; GWAS ; nested mixture model

Abstract

Genome-wide association studies (GWAS) have rapidly become a standard method for disease gene discovery. To overcome the limitations of single-SNP analysis, there have been increasing efforts recently on GWAS pathway analysis, aiming at combining SNPs with moderate signals. However, a major drawback of current pathway-based methods is that interactions among genes within a pathway are ignored and genes are treated as exchangeable, leading to inefficient use of biological prior knowledge and loss of power. Here we propose a flexible Bayesian nested mixture model to incorporate genome-wide gene-gene interaction information embedded in gene networks into gene-based analysis of GWAS data. We carry out parameter estimation and inference based on MCMC samples in a Bayesian variable selection framework. Applications to real GWAS datasets, together with simulation studies, demonstrates the extra power gained by integrating gene networks with GWAS data.