In the present samples, the association between rs73568641-C and higher opioid dose was observed only in AAs for both methadone and morphine. In our prior OD GWAS,24 which included most of the Yale-Penn subjects in the present study, we similarly reported very different results for AAs and EAs, with the most significant results (which did not include any markers near OPRM1) in AAs. In the present study the GS association signal was stronger in methadone-treated AAs than EAs. There are several possible explanations that could account for associations being preferentially detected in specific populations. GWAS SNPs often tag many common variants, as is the case here (Figure 1), and population-specific GWAS findings63 may be related to linkage disequilibrium between commons SNPs and population-specific rare functional variants.64 Whole genome sequencing approaches and larger samples will be needed to interrogate fully variation across the allele frequency spectrum in multiple ancestry groups. Epistasis provides another possible explanation; some polymorphisms may have phenotypic effects only when population-specific variants in the region or even elsewhere in the genome are present to interact with them.65
