Abstract 1296: Bayesian fine-mapping using summary data of 145,000 subjects refines common risk associations, discovers secondary signals and novel candidate genes for prostate cancer

Abstract

Abstract Genome-wide association studies (GWAS) have identified more than 160 prostate cancer (PrCa) genetic risk loci, however these variants rarely point directly to the true underlying functional variant driving the association. In this fine-mapping study to narrow the credible causal variant set for 80 PrCa regions representing 89 original independent GWAS signals, we performed Bayesian variable selection in combination with functional annotation and quantile regression. We used imputed data for 83,511 PrCa cases and 62,283 controls investigated with high-density genotyping arrays from the OncoArray, iCOGS and 5 previous GWAS studies from the PRACTICAL/ELLIPSE consortia. To facilitate fine-mapping from one-at-a-time SNP associations meta-analyzed over the consortia we first applied JAM, a novel Bayesian algorithm which searches multi-SNP models in summary data by imputing the correlation structure according to a reference panel. JAM provides inference on the number of independent signals, as well as the set of potential SNPs driving those signals. We utilized functional annotation and eQTL analysis (TCGA prostate tumor data) in combination with quantile regression to further prioritize the most likely causal variants within the credible set of SNPs and identify potential candidate genes and functional mechanisms. The median credible set size from JAM was 17 SNPs per region, shrinking the post-QC input set of variants by about 98%. In 13 regions evidence was found for multiple independent signals, up to a maximum of 5 SNPs. Within the single hit regions, almost half had less than 10 variants selected. In 34 regions the credible set included at least one SNP that was co-localized with a significant eQTL. Quantile regression highlighted enrichment for variants in promoters, DNase hypersensitivity site and eQTLs - representing candidate biological mechanisms underpinning disease development. This study has substantially reduced and prioritized the candidate causal PrCa risk variants within previously known GWAS regions, identifying a small subset of variants for further functional investigation and novel candidate genes at a number of loci. Citation Format: Zsofia Kote-Jarai, Tokhir Dadaev, Ed Saunders, Paul Newcombe, Ezequiel Anokian, Daniel Leongamornlert, Ali Amin Al Olama, Christopher Haiman, Ros Eeles, David Conti, The PRACTICAL/ELLIPSE Consortium. Bayesian fine-mapping using summary data of 145,000 subjects refines common risk associations, discovers secondary signals and novel candidate genes for prostate cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 1296. doi:10.1158/1538-7445.AM2017-1296