Imputation using IMPUTE2 [17] was preceded by prephasing the study genotypes with the ShapeIT program to estimate haplotypes, using 500 conditioning states, an effective population size of 17,469 as recommended for populations of African descent, and default settings for all other program options. The estimated study haplotypes were then input into IMPUTE2 to impute SNPs available on the reference haplotype panel. Imputations were done on 4.5 MB chunks with 1 MB flanking buffers. Default options were used, except that k_hap was set to the number of haplotypes in each reference panel. Smaller values of k_hap would likely produce similar accuracy at lower computational cost [17].
