Extended Data Fig. 6

Global population structure. Results of the ADMIXTURE algorithm and hierarchical clustering of outgroup F₃-statistic values for four continental samples from the 1000 Genomes collection; A. Europe, B. East Asia, C. Africa and D. South Asia. The maps depict the sampling locations for each population, in addition to the sample size used (n = 28 per population). Note that approximate locations for some populations (i.e. CEU, ITU and STU) are given as per the original 1000 Genomes publication¹⁴. Coloured tip-points below each leaf of the hierarchical clustering tree depict the geographic population label of the individual (from the maps). Hierarchical clustering was not performed on African samples due to the use of Yoruba as the outgroup population (see Methods). The bar charts show the output of the clustering algorithm ADMIXTURE applied to each sample, assuming the same number of clusters as the geographically defined samples (K = 5 for Europe, East Asia and South Asia, and K = 7 for Africa). Clustering makes no reference to the sampling locations of the individuals and is based on genetic data alone. Individuals are listed along the x-axis, grouped according to their sampling location, with bars above reflecting their cluster assignment in the following manner: each inferred cluster is labelled by a colour and the proportion of bar assigned that colour represents the probability that the individual is assigned to that cluster. Colours were manually selected (post-hoc) to match the labels in the maps. See Fig. 2 of the main text for the results of these same algorithms when applied to the NCIG + PNG dataset.