Although these findings were obtained from a limited sampling of African genetic diversity, they provide some intuition about general strategies for imputation in people of recent African ancestry. It will always help to collect new reference data through population-specific sequencing or genotyping, but the gains from this approach will often be larger in Africa. Whether or not a well-matched panel is available for a particular study population, imputing from all available haplotypes of African ancestry may often improve the results. Non-specific reference panels can weaken association signals near loci that have recently experienced strong selection, as with the beta-globin region in the Jallow et al. study, so it may be worthwhile to re-impute from just the ancestry-matched reference haplotypes (when such haplotypes are available) in regions showing decisive evidence of selective sweeps.
