Identification of novel type 1 and type 2 diabetes genes by co-localization of human islet eQTL and GWAS variants with colocRedRibbon

Over 1,000 distinct genetic variants have been associated with diabetes risk by genome-wide association studies (GWAS) but for most their functional impact is unknown and less than 15% of the diabetes GWAS variants have been shown in expression quantitative trait locus (eQTL) studies to alter gene expression in pancreatic islets. To fill this gap, we developed a new co-localization pipeline, called colocRedRibbon, that prefilters eQTL variants by direction of effect on gene expression, shortlists overlapping eQTL and GWAS variants and then runs the co-localization. Applying colocRedRibbon to diabetes and glycemic trait GWAS, we identified 292 co-localizing gene regions - 236 of which are new - including 24 co-localizations for type 1 diabetes and 268 for type 2 diabetes and glycemic traits. We achieved a four-fold increase in co-localizations, with the novel pipeline and updated GWAS each contributing two-fold. Among the co-localizations are a low frequency variant increasing MYO5C expression that reduces type 2 diabetes risk and a type 1 diabetes protective variant that increases FUT2 and decreases RASIP1 expression. These novel co-localizations represent a significant step forward to understand polygenic diabetes genetics and its impact on human islet gene expression.