of the East Asian log(OR). These figures would be unexpected if GWAS hits were synthetically generated by population-specific rare causal variants, as their effect size and Linkage Disequilibrium (LD) with the replicated SNP would be different in each population. Moreover, when considering only replication attempts that did not achieve P<0.05 in East Asians, there is still a strong and significant correlation between the two ancestries' OR (Spearman's ρ = 0.53, P<2·10−9), which suggests that a fraction of these associations might also be shared even if not successfully replicated in East Asians. A final further piece of evidence indicates that causal alleles behind non-replicated SNPs might actually be common and shared. Since most rare variants occurred after the split of Europeans and East Asians [4], [12], [26]–[28], they would have accumulated randomly in the genealogy of each allele of the tagSNP used in GWAS. Therefore, if causal variants were rare, risk alleles would not be necessarily shared even if discovered through the same tagSNP. Strikingly, when considering the direction of effects of SNPs non-replicated in East Asians instead of only their significance, the same risk allele as in Europeans was observed for 73.6% of attempts. This proportion clearly departs from the
