Asian) fulfilled our criteria (average ∼10 papers per disease, range = 3–27 papers; see Materials and Methods and Tables S1 and S2). We further performed an exhaustive search of the literature to detect any GWAS published before February 2012 but not available in the Catalog (see Materials and Methods). This effort yielded six new GWAS that we included in our database (Table S3 and Table S2, in yellow). Finally, and to be as comprehensive as possible, we included 16 GWAS performed on peoples of African American ancestry that were available for eight of the previously ascertained 28 diseases. This rendered a final dataset of 299 GWAS papers reporting 419 associations to 28 diseases (Table S2). Out of these, we ascertained 190 SNPs initially reported as genome-wide significant (P<5×10−7, or P<5×10−8 if the study included imputed SNPs) in European GWAS and for which one or more replication attempts had been performed in subsequent European, East Asian and/or African GWAS (181, 225 and 61 attempts, respectively, Tables S4 and S5). We studied patterns of replication across studies, using the criterion that a replication was successful if the same risk allele achieved P<0.05. To obtain that information we examined every individual paper, since
