Regions of the genome exhibiting moderate to strong evidence for CD, RA and T1D risk factors.

For all CD and RA loci in this table, there is at least a 0.5 probability that one or more SNPs in the region are included in the multi-marker disease model (); for all T1D loci, . Support for disease associations is conditioned on enrichment of pathways in Figure 1. Rows marked with * are selected only after accounting for pathway enrichment, or show substantial increase in support due to feedback from enrichment. Right-most column cites published GWAS findings that corroborate majority of * rows. In this column, ** indicates that validation is not required as disease association is already strongly supported without pathways; these rows recapitulate the strongest associations reported in the original study [65] (see Supplementary Materials). Genes in enriched pathways are written in bold. Table columns from left to right are: (1) disease; (2) chromosomal locus; (3) region most likely containing the risk-conferring variant(s), in Megabases (Mb); (4) posterior probability that one or more SNPs in region are included in model under null, and (5) under enrichment hypothesis; (6) posterior probability that two or more SNPs are included under null, and (7) under enrichment hypothesis; (8) smallest trend p-value in region from original analysis [65], when available (some of these p-values are derived from imputed SNPs, and are not available in our data); (9) established genes in disease pathogenesis, or most credible genes of interest based on prior studies, corresponding to locus (when the most credible gene differs from gene assigned to pathway, pathway gene is shown in parentheses); (10) refSNP identifier of SNP in critical region with largest PIP (this SNP is likely in linkage disequilibrium with the causal variant rather than being causal itself, and may not match SNP reported in [65] with smallest p-value); (11) the PIP of this SNP; (12) posterior mean of log-odds ratio (additive effect of minor allele count on log-odds of disease) given SNP that is included in multi-marker disease model; (13) 95% credible interval of effect size, ; (14) frequency of minor allele for SNP in controls, and (15) in cases. Bold numbers in and columns highlight appreciable increase in support for disease associations within region after feedback from enriched pathway. Credible interval is smallest interval about posterior mean that contains with 95% posterior probability. The “critical region” at each locus is estimated by inspecting single-SNP BFs [79], and bounding the region by areas of high recombination rate, inferred using data from Phase I, release 16a of the HapMap study [185], and visualized in UCSC Genome Browser [186]. Note that statistic for critical region may be slightly different than for overlapping segment shown in Figure 3 due to different numbers of SNPs in segments and critical regions. All SNP information and genomic positions are based on Human Genome Assembly 17 (NCBI build 35).