B0232
Title: Bayesian GWAS with structured and non-local priors
Authors: Adam Kaplan - University of Minnesota Twin Cities (United States) [presenting]
Mark Fiecas - University of Minnesota (United States)
Eric Lock - University of Minnesota (United States)
Abstract: A novel Bayesian approach to genome-wide association studies (GWAS) is introduced which improves over existing methods in two important ways. First, we describe a model that allows for a marker's gene-parent membership and other characteristics to influence its probability of association with an outcome. For this we use a hierarchical Dirichlet Process (DP) model that allows for clustering of the genes in tandem with a regression model for marker-level covariates. Second, we use Non-Local priors to model the difference in probability of minor allele status between patient disease status. We outline the implementation of and discuss the philosophical problems treated by Non-Local priors within the genome-wide analysis framework. In Bayesian hypothesis testing, it is often overlooked that the null hypothesis is a sub-event of the alternative hypothesis. This results in the asymptotic rates of convergence favoring the alternative hypothesis over the null, whereas we define a Non-Local prior for the GWAS context that gives symmetric rates of convergence. We assess the structured and Non-Local components with simulation studies under various scenarios. We apply our Bayesian GWAS method to single-nucleotide polymorphisms (SNP) data collected from a pool of Alzheimer's disease and cognitively normal patients from the Alzheimer's Database Neuroimaging Initiative.
