Finally, we considered replication attempts performed upon individuals of African ancestry. Even if GWAS on individuals of non-Eurasian ancestry are scarce, we could find 16 GWAS performed on African Americans from which we gathered a total of 73 replication attempts (61 and 12 for SNPs discovered in Europeans and East Asians, respectively; see Materials and Methods and Table S1). Overall, we observed a low replicability rate (9.6%, 7 out of 73 attempts) that was not attributable to lack of statistical power (59.2% on average, see Table S5). This figure would cast doubts about the sharing of causal variants between Eurasians and Africans, but the inherent limitations of this part of the analysis warrant for caution. For instance, lower levels of LD in Africans than in Eurasians make it difficult to ensure that potentially shared causal variants are tagged by the same marker SNP [25]. Additionally, the 16 African-American GWAS form a rather small dataset corresponding to only five diseases (asthma, cardiovascular disease, hypertension, prostate cancer and type 2 diabetes). A complete study of African replicabilities will be possible when more studies are available. In the meantime, we focused on data gathered from European and East Asian GWAS.
