Bayesian Hierarchical Models with Spatial Priors in Genome-Wide Association Studies

• Activity Number: 657
• Type: Contributed
• Date/Time: Thursday, August 4, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Jie Shen*+ and Hal Stern
• Companies: University of California at Irvine and University of California at Irvine
• Keywords: Bayesian hierarchical modeling ; GWAS ; spatial models ; dependence

Abstract

Genome-wide analyses are now commonplace approaches to identifying genetic risk factors for disease. The standard analysis is based on a series of SNP tests for association and ignores known relationships among SNPs. This motivates our work on methods that incorporate dependence among markers. In this study, we develop Bayesian hierarchical models to analyze a range of SNPs simultaneously. Our basic approach is similar to the traditional approach, in that it assumes a test statistic computed separately at each SNP, which is presumed to be a noisy version of the underlying effect. Two types of prior probability models are put on the underlying effects to capture dependence among SNPs: a model allowing for dependence as a function of distance between SNPs and a model integrating dependence among multiple markers within a gene. We illustrate the models on simulated data and on a real Alzheimer's disease dataset, which confirm that our models have higher power of statistical inference.