Figure 1

Scree Plot and Heatmap of Top 20 PCs: The scree plots and heatmaps are presented for the simulated data with no structure (left), HGDP Africans (middle), NECS (right). The scree plot graphs the natural logarithm of the eigenvalue (y-axis) versus the component number (x-axis). A kink is not observed in the scree plot for the simulated data with no population substructure (a). For the HGDP Africans (b) and the NECS (c) the scree plot identifies a kink at the 5^th and 4^th PCs, respectively. In the heatmaps, each column is a PC and each row is an individual. Original PCs are standardized by row and the color intensities correspond to the standardized value of the PC for each individual (green: higher than average, red: lower average) and are sorted by the corresponding eigenvalues, and rows are sorted by hierarchical clustering. While no pattern is found in the heatmap of the simulated example without population structure, a distinct pattern is observed for the HGDP Africans and the NECS. The pattern in the HGDP Africans is the most distinct since these populations are well defined and are very different from each other. The pattern for the NECS is more subtle because the variability in subjects of European ancestry is much lower than in Africans. However, one can still observe variability beyond the first 2 principal components.