To more closely examine the extent of sample structure within the NFBC66, we used PCA of the genotype covariance matrix9 and multidimensional scaling analysis (MDS) of the identity-by-state (IBS) matrix from NFBC66 samples. The first two coordinates identified by MDS are known to correlate well with the geographical location of the linguistic groups13. The first two principal components in the current sample correlate well with latitude and longitude of parental birthplaces for the subset of individuals with known ancestry (Fig. 1). Indeed, we noted that PCA of genotypes and classical MDS of the IBS matrix lead to very similar results. There is a correlation coefficient of 0.9993 between the first components from PCA and MDS and a correlation coefficient of 0.9978 between the second components. The first five principal components separate to varying degrees the linguistic and geographic subgroups comprising northern Finland (Supplementary Fig. 1), consistent with the previous analysis using MDS13. Despite the clear correlation between geographical regions of origin and the first two principal components, clustering analyses of the IBS matrix using PLINK software or hierarchical clustering in R did not identify separate subgroups.
