The total fraction of variance explained was calculated using the formula 2f(1−f)*a2, where f is the frequency of the variant and a is the additive effect of the variant9. When calculating percent variance explained in the African ancestry sample, for the previously-discovered BMI variants that were not genome-wide significant in Stage 1, we used data from the Stage 1 sample; for those that were genome-wide significant we used data from the Stage 2 sample; and for the novel BMI variants we used data from the Stage 2+3 samples to avoid inflating the estimates due to the winner’s curse. When summing percent variance explained for the 36 previously-discovered BMI variants (Supplemental Table 12), we utilized the more informative SNP discovered through fine-mapping at the seven loci (listed in Supplemental Table 13). However, for these seven variants Stage 1 results were used and estimates may be biased; Stage 2 and 3 studies only participated in the look-up of the top SNPs from preceding Stages.
