Fig. 6

Estimation of pangenome properties from incomplete genomes. a Comparison between pangenome properties inferred from high-quality genomes (x-axis) and mixtures of medium-quality MAG and high-quality genomes (y-axis). From left to right, the scatter plots correspond to pangenomes with 5, 20, 50, and 100% of MAG. Each point in the scatter plots corresponds to one pangenome, with different symbols and colors used to distinguish among species and OGC generation methods, respectively. Each scatter plot combines data from 4 species, 6 gene clustering methods, and 3 random subsamples. The bar plots on the right summarize the observed inconsistencies, with color intensities representing the fraction of MAG. Note that panX fails to produce results in pangenomes that contain > 5% of MAG (purple asterisks). b Sensitivity of gene clustering methods to the addition of MAGs, calculated by comparing the cluster assignations of genes from high-quality genomes before and after adding MAG. NVI: normalized variation of information. Different OGC generation methods are color coded, with color intensities indicating the fraction of MAG. Note that reference-database mapping methods produce an NVI equal to zero (double asterisk)