To characterize alleles that might better represent the biologically functional variant at the 36 previously-discovered BMI loci, we searched for LD proxies among individuals of African ancestry. Using HapMap data (CEU or JPT/CHB) to estimate LD, we identified all SNPs that were correlated (r2≥0.4) with the index SNP (within 250 kb, or larger to include a nearby gene). Next, we tested these SNPs for association with BMI in the Stage 1 African ancestry sample. We applied a locus-specific significance criterion α, which accounts for multiple testing [the number of tag SNPs in the HapMap YRI population that capture (r2≥0.8) all common SNPs (MAF ≥0.05) correlated with the index signal in the HapMap CEU or JPT/CHB populations]. This alpha level does not account for the number of regions evaluated and reflects a balance between the need to correct for multiple comparisons and the prior knowledge that each region harbors a risk variant for BMI. We also looked for novel independent associations, focusing on the genotyped and imputed SNPs that were uncorrelated with the index signal in the initial GWAS populations (r2<0.2).
