Figure 2.

Robustness against batch artefacts during the projection of two distinct HGDP data partitioned samples, one to modify (pink) and one to remain unchanged (blue), onto a 1KGP based reference ancestry space. First Column, (A), (D), and (G), no real modification of the genotypes was done whereby both data partitions, pink and blue, overlap perfectly for PCA, UPCA and SUGIBS, respectively. Second column, (B), (E), and (H), a first actual modification was done by introducing missing genotypes in the data of the partition to modify (pink) with results for PCA, UPCA and SUGIBS, respectively. Third column, (C), (F), and (I), a second actual modification was done by modifying genotypes in the data of the partition to modify (pink) with results for PCA, UPCA and SUGIBS, respectively.