Fig. 5

VNTR concordance in three pangenomes. a TRs with identical counts in at least two pangenomes, with the median count for the cattle and non-cattle groups. A particular VNTR with substantially more repeats in non-cattle compared to cattle that we investigated further is circled in red. b Pearson and Spearman squared correlation coefficients across the TR counts for pggb-cactus (p–c), pggb-minigraph (p-m), and cactus-minigraph (c-m). Each point is one assembly, with box plots over the 12 assemblies. c Similar to a, except TRs with identical counts in pggb and cactus that were not present in minigraph. d adVNTR-derived genotypes for five HiFi samples in the three pangenomes. The black dashed line indicates the expected count using adVNTR as a ground truth. e VNTRs where all three pangenomes agreed to a different count than adVNTR, suggesting adVNTR may sometimes over-/underestimate assembly-based counts. f Trees derived from the TR counts across different input assemblies, with colours representing clusters of taurine cattle, indicine cattle, and non-cattle