To evaluate the use of dense genotype data to predict geographic ancestry in the Icelandic population, we randomly selected 250 additional Icelandic samples for which genotype data was available (see Materials and Methods). A PCA run with the 250 samples included (Figure 2B) indicates that these individuals trace their ancestry from all over Iceland, with an excess of individuals from the vicinity of region 4 (concordant with Table 1). We used the PCA results to predict the regional ancestry of each of the 250 samples and compared this with their true ancestry, which we defined as the region in which the greatest number of ancestors five generations back was born (see Materials and Methods). The ancestry predictions were correct for 47% of samples, correct to within a distance of one region for 74% of samples, and correct to within a distance of two regions for 93% of samples. The accuracy increased to 58% (87% to within one region, 97% to within two regions) when restricting to the 98 (of 250) samples with at least 16 of 32 ancestors from a
