To discover type 2 diabetes (T2D) loci and enhance fine-mapping resolution, we conducted the largest meta-analysis of genome-wide association studies of the disease to date by aggregating 171,262 cases and 1,075,072 controls from diverse populations (45% non-European ancestry). We identified 250 loci at genome-wide significance (p<5x10-8); 59 mapped outside regions previously implicated in T2D. Across these loci, conditional analyses revealed a total of 475 distinct signals of association (locus-wide significance, p<10-5). We observed strong evidence of heterogeneity in allelic effects on T2D (pHET<1.4x10-4, Bonferroni correction) correlated with ancestry at 17% of signals, including LEP (rs35589574, pHET=4.8x10-25, East Asian specific) and multiple associations at/near KCNQ1 and TCF7L2 (representing ethnic-specific/-differentiated effects). T2D-associated variants showed significant enrichment (odds-ratio range 1.90-6.63; p<0.05) in coding exons, pancreatic islet enhancers and promoters, adipose enhancers, and binding sites for transcription factors, including NKX2.2 and FOXA2. Increased sample size, population diversity, and annotation-informed fine-mapping substantially improved localisation of potential causal variants compared with previous efforts, and highlighted 76 signals with a single variant accounting for >80% of the posterior probability of association (PPA); of these 35 signals had PPA of >99%. Clustering of signal-specific annotation enrichment highlighted distinct clades of T2D associations driven by different underlying molecular processes. These analyses represent the most comprehensive view of the genetic contribution to T2D to date and, through integration with expression quantitative trait loci in disease-relevant tissues, point to previously unreported effector genes (such as SKOR1, CLUAP1, and PEPD) and mediating molecular mechanisms at several loci.
H. Kitajima: None. X. Sim: None. M. Ng: None. W. Zhang: None. J.E. Below: None. A.J. Payne: None. K. Gaulton: None. A. Morris: None.