Fine-mapping gene-based associations via knockoff analysis of biobank-scale data with applications to UK Biobank

We propose BIGKnock (BIobank-scale Gene-based association test via Knockoffs), a gene-based testing approach that leverages long-range chromatin interaction data, is applicable to biobank-scale data, and performs conditional testing genome-wide via knockoffs. Thereby BIGKnock reduces the confounding effect of linkage disequilibrium and can prioritize causal genes over proxy associations at a locus. We applied BIGKnock to the UK Biobank data with 405,296 British subjects for multiple binary and quantitative traits, and show that relative to conventional gene-based tests BIGKnock produces smaller sets of significant genes that contain the causal gene(s) with high probability. We further illustrate its ability to pinpoint potentially causal genes at ∼ 80% of the associated loci (4,829 loci across 24 diseases and traits), including genes with well established causal links in the literature such as ASGR1 and ANGPTL4 and cholesterol, and ALDH2 and coronary artery disease, as well as plausible novel links such as NGFR and asthma, AGPAT1 and type 2 diabetes, DBH and blood pressure, ZHX3 and calcium, PPARγ and LDL cholesterol. Relative to several methods for causal gene prioritization such as closest gene, cS2G and L2G, we show that BIGKnock produces more interpretable results and can help improve precision on a set of gold standard causal genes.