To assess whether our imputation strategy would yield similar benefits in African populations outside the HapMap 3 set, we performed additional cross-validations in a Gambian dataset from the Malaria Genomic Epidemiology Network (MalariaGEN; Malaria Genomic Epidemiology Network 2008). Previous work on imputing Gambians in a disease study found that the HapMap 2 YRI panel produced weaker association signals than did a Gambia-specific panel at the strongly selected beta-globin gene (Jallow et al. 2009). Here, we extend that work to examine imputation accuracy in larger reference panels and a broader variety of loci. We also rephrase the question to ask whether haplotypes sampled outside The Gambia can improve accuracy when a dedicated Gambian panel is available.
