Properly controlling for Type I errors in large-scale hypothesis testing of genetic studies has been a challenge given currently obtainable sample sizes. Bonferroni-derived thresholds are severely underpowered to detect the vast majority of associations. Local false discovery rate (fdr) methods provide more power to detect non-null associations, but implicit assumptions about exchangeability and independence limit their ability to discover non-null hypotheses. We leverage the information available in genetic annotation categories for Genome-wide association studies (GWAS) and biological pathways, such as KEGG (Kyoto Encyclopedia of Genes and Genomes). To this end, we develop a novel Bayesian approach for generalized covariate-modulated local false discovery rate (cmfdr) estimation. The proposed methods releases us from both assumptions of exchangeability and independence. Through a vast array of simulation experiments and applications to Crohn’s disease and Schizophrenia GWAS, we demonstrate notable power gain and increased sensitivity while controlling the total false discovery proportion at the desired level.
