Validation of Ancestry Estimates

Robust and Consistent Ancestry Estimates
Estimates from Ancestry Composition are extremely well calibrated, with correlations of African, European, and Native American ancestry estimates showing r2 > 0.98 with 1000 Genomes Project African American and Latino consensus estimates (Figure 5 from Durand et al.33). Admixture tests via an independent admixture software package, ADMIXTOOLS,64 confirm significant signals of African admixture in European Americans (Table S7). Ancestry Composition estimates are highly concordant with ADMIXTURE66 estimates, with r2 values of 0.94, 0.98, and 0.91, for the three groups, respectively (Figure S14).

Evidence that the Great Majority of Ancestry Segments that We Detect Are Real
We show that positions of segments of non-European ancestry start uniformly across the genome (see Figure S15). Although some regions, including the HLA region containing the MHC complex on chromosome 6, show higher ancestry switches reflecting difficulties in assignment because of genetic diversity (as likewise seen in African Americans and Latinos; Figures S16 and S17), the majority of segments are uniformly distributed. Only 4% of all segment starts of African ancestry lie within the HLA region, and only about 1.4% of Native American segment starts lie in the HLA region.
We find very low levels of African and Native American ancestry in Europeans with four grandparents born in Europe. We estimate that only 0.98% of Europeans carry African ancestry and 0.26% of Europeans carry Native American ancestry. These levels are substantially lower than the 3.5% and 2.7% of European Americans who carry African and Native American ancestry, respectively. Furthermore, for most European countries we observed no individuals with substantial non-European ancestry, and the presence of individuals with African and Native American ancestry is limited to countries that had major ports in the Atlantic trade and were known to have been highly connected to the trans-Atlantic slave trade. Indeed, African ancestry in individuals from Europe is not unexpected; approximately 9,000 Africans were brought to Europe between 1501 and 1867 (as documented by Eltis and Richardson’s maps of the slave trade, accessible at Emory University’s database). Excluding countries that had major and minor ports in the Atlantic with strong connections to the slave trade (namely Portugal, Spain, France, and United Kingdom) and Malta, which has been the site of migrations from Africa and the Middle East, we obtain a data set of 9,701 Europeans, where we find African and Native American ancestry is virtually absent, with only 0.04% of individuals carrying 1% or more African ancestry and 0.01% carrying 1% or more Native American ancestry, within the margins of survey error estimates.

Native American mtDNA in European Americans and African Americans and Not in Europeans
The frequency of Native American mtDNA haplogroups in European Americans and African Americans correlate with our estimates of genome-wide ancestry in European Americans and African Americans and are found in appreciable fractions of individuals who are estimated to carry Native American ancestry. The frequencies of haplogroups are shown in Table S8. These haplogroups are virtually absent in individuals with four grandparents from a European country (21 individuals out of 15,651). Furthermore, the majority of these Native American haplogroups in Europeans are found in individuals from Spain. Though it is possible these represent non-Native American haplogroups, prior literature and studies of genetic, archaeological, and paleontological evidence suggest that these haplogroups have Native American origins and is evidence of gene flow from the Americas to Spain. Excluding Spain, Native-American-specific haplogroups are detected in fewer than 0.05% of individuals with four grandparents from Europe and can be explained by survey errors in reporting all four grandparents’ birth places.