Principal components analysis (PCA) is a widely used tool for analyzing genetic data [15]–[18]. We ran PCA on genotype data from the 877 individuals using the EIGENSOFT software with default parameters settings [17]. A plot of the top two principal components is displayed in Figure 2A, revealing a striking concordance between the geographical orientation of the 11 regions (Figure 1) and the relative positions of each region on the PCA plot (Figure 2A). In both cases, we observe a ring-shaped topology with region numbers increasing in clockwise order and a central void corresponding to the unpopulated interior of Iceland. The top two PCs explain a modest proportion of the overall variance: 0.0027 for PC1 and 0.0022 for PC2, representing an excess of 0.0015 for PC1 and 0.0011 for PC2 above what would be expected by chance (Tracy-Widom P-values<10−12 in each case [17]), similar to previous results on European American data sets [13]. We note that these PCs are the result of genome-wide structure, as opposed to a small number of highly informative markers (see Text S1).
