Analysis was performed without any prior population assignment using STRUCTURE [29] (see Methods). The European ancestry groups, western European American (WEURA), central European American (CEURA), eastern European American (EEURA), Finland (FND), Italy (ITN), Spain (SPN), Sweden (SWD), and other European American (OEURA) are indicated by color code. The latter group consisted of individuals with mixed European ancestry from several regions and additional smaller groups (see Methods). The average contribution of each color-coded cluster is indicated by the proportion of the horizontal bars in (A), (B), and (C), whereas in (D), the proportion of each cluster (ordinate) is shown for each individual.
(A) Analyses were performed with 2,657 SNPs under the condition of two population (Pop) groups (k = 2).
(B) An analysis is shown for k = 10.
(C) The results of an analysis using only a selected subset of 400 SNPs is shown for k = 2. These SNPs were selected for potential informativeness for the north–south division in European population structure (see text). None of the participant samples used for selecting this SNP panel (including the Spanish samples) were included in this analysis assessing these same markers.
Panel (D) Depicts individual participant results under same conditions as (A).

For each self-identified group, the fraction of individual participants in each group with the color-coded frequency “southern” contribution is shown. For southern European American I (SEA1), only southern European grandparents were identified. For southern European American II (SEA2), grandparents were self-reported as being of both southern European decent and western, central, or eastern European decent. For the eastern Mediterranean American (EMEDA) group, four of ten were of mixed-European decent with one or more grandparents of western, central, or eastern European decent.
CEURA, central European American; EEURA, eastern European American; FND, Finland; ITN, Italy; PTG, Portugal; SCNA, Scandinavian; SJA, Sephardic Jewish American; SPN, Spain; SWD, Sweden; WEURA, western European American; WFC, White French Canadian.

Analysis was performed using 749 SNPs informative for European substructure using STRUCTURE. These summary results show the percentage of individual participants in each group with the color-coded percentage “southern” contribution. (A) and (B) show results from participants recruited in New York City as part of the New York Cancer Project.
For (A), the individuals were grouped by regional location based on available grandparental data from one or more grandparents.
(B) A subset of the same participants as in (A), those with four grandparents from the same country of origin, is shown.
(C) shows the results of a different set of participants chosen or recruited specifically on the basis of additional ethnic information (Jewish and eastern European descent). Each of the Ukrainian participants did not have known Jewish ancestry. Similarly, the two participants of mixed Ukrainian and Russian ethnic ancestry (two grandparents Ukrainian, and two grandparents without known Jewish ancestry. For the Ashkenazi Jewish participants, the eastern European group had varying countries of origin, including Ukrainian, Polish, Lithuanian, Russian, and Romanian. For the mixed central and western European Ashkenazi, the participants included those with one to three grandparents from Germany, Austria, or Hungry (central), or England or Belgium (west), and other grandparents from eastern European countries. Each of these Ashkenazi Jewish participants self-reported four grandparents of Ashkenazi Jewish ancestry.
CEURA, central European American; EEURA, eastern European American; EMEDA, eastern Mediterranean American; JEA, Jewish ancestry; MEA, Mediterranean European American; SCNA, Scandinavian; SEURA, southern European American; WEURA, western European American.

The individual participants are represented by rectangular shapes distributed by the strength of their separation in three dimensions: along the first factor (abscissa), second factor (ordinate), and third factor (depth). This factor analysis is based on vectors fitted to the individual allele frequencies of each SNP. The percentage of inertia for each factor is provided on each of the axes and correspond to the eigenvalue vectors.
(A) The analysis utilized the set of 2,657 random SNPs. In the upper panel, the European ancestry groups, Italy (ITN), Spain (SPN), eastern European American (EEURA), western European American (WEURA), Sweden (SWD), and Amerindian (AMI) are indicated by color code, and show that the ITN and SPN participants are mostly distinct from the other European populations examined. The eigenvalue vectors for factors 1–3 were 0.0234, 0.0024, and 0.0019, respectively. In the bottom panel, the same groups are shown without the AMI participants. The eigenvalue vectors for factors 1–3 were 0.0033, 0.0022, and 0.0014, respectively.
(B) The analysis was performed using 749 SNPs chosen for European substructure information. The country of origin is shown by the color coding indicated in the upper panel. Except for the Spanish, Italian (1), and Swedish groups, the participants in (B) do not overlap with those in (A), and were European Americans self-reported as having four grandparents with the same country of origin. The Italian (2) group were Italian Americans. The bottom right of (B) shows a different three-dimensional view (factor 3 as ordinate), and the bottom left of (B) shows factor 4. The vector eigenvalues for factors 1–4 were 0.0059, 0.0026, 0.0017, and 0.0016, respectively.

The model examined the real genotypes of Italian participants compared with western and northern European participants, for three different SNPs indicated on the abscissa. The log of 1 divided by probability is shown on the ordinate with the nominal p-values (association test without controlling for population structure) indicated by the open diamonds, and the results of the structured association with the entire set of 2,657 SNPs depicted with the gray triangles. The results of the structured association tests are shown for ten random sets of 400 SNPs (filled circles) and for the set of 400 SNPs chosen for north–south population structure (see text) (gray squares).