Estimating ancestral proportions in a multi-ethnic US sample: implications for studies of admixed populations

This study was designed to determine the ancestral composition of a multi-ethnic sample collected for studies of drug addictions in New York City and Las Vegas, and to examine the reliability of self-identified ethnicity and three-generation family history data. Ancestry biographical scores for seven clusters corresponding to world major geographical regions were obtained using STRUCTURE, based on genotypes of 168 ancestry informative markers (AIMs), for a sample of 1,291 African Americans (AA), European Americans (EA), and Hispanic Americans (HA) along with data from 1,051 HGDP-CEPH 'diversity panel' as a reference. Self-identified ethnicity and family history data, obtained in an interview, were accurate in identifying the individual major ancestry in the AA and the EA samples (approximately 99% and 95%, respectively) but were not useful for the HA sample and could not predict the extent of admixture in any group. The mean proportions of the combined clusters corresponding to European and Middle Eastern populations in the AA sample, revealed by AIMs analysis, were 0.13. The HA subjects, predominantly Puerto Ricans, showed a highly variable hybrid contribution pattern of clusters corresponding to Europe (0.27), Middle East (0.27), Africa (0.20), and Central Asia (0.14). The effect of admixture on allele frequencies is demonstrated for two single-nucleotide polymorphisms (118A > G, 17 C > T) of the mu opioid receptor gene (OPRM1). This study reiterates the importance of AIMs in defining ancestry, especially in admixed populations.