Model Selection v. Multiple Testing in Genome-Wide Association Studies

• Activity Number: 610
• Type: Topic Contributed
• Date/Time: Thursday, August 4, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Florian Frommlet*+ and Felix Ruhaltinger and Malgorzata Bogdan
• Companies: Medical University Vienna and Medical University Vienna and TU Wroclaw
• Keywords: Model selection ; Multiple Testing ; Genome Wide Association Studies

Abstract

For the vast majority of genome wide association studies (GWAS) published so far, statistical analysis was performed by testing markers individually and then correct for multiple testing. Here we will compare such a multiple testing approach with multiple regression for quantitative traits and generalized linear models for case control studies. We introduce a model selection approach to GWAS based on modifications of Bayesian Information Criterion (BIC) which control either the family wise error (mBIC) or the false discovery rate (mBIC2). Simple statistical considerations suggest that model selection has larger power than multiple testing to detect causal SNPs in complex models. This is confirmed by comprehensive simulations based on real SNP data. On the other hand we show that multiple testing has substantial problems with proper ranking of causal SNPs and is prone to detect a certain number of false positive SNPs, which are not linked to any of the causal mutations.