Fig. 3.

Population structure of large indels at three different levels. a Super-population level: the average ratio of indels identified from samples in each super-population that are specific to that super-population, shared with some other super-populations but not all, or shared with all other super-populations. Random sub-sampling has been applied to balance the sizes of super-populations. The reported values are the average of 100 random sub-samples. b Population level: a phylogenetic tree constructed based on the indel occurrence matrix. c Single-sample level: the first two principal components of the indel occurrence matrix based on super-population groups. AFR Africans, AMR Americans, EAS East Asians, EUR Europeans, SAS South Asians