Fig. 2

Method-dependent variation and uncertainty in pangenome features. a Consensus similarity tree of OGC building methods based on the species-wise normalized variation of information for the assignation of ORF to OGC. Labels indicate the number of species (out of 125) that support each branch. b Consensus similarity tree of different pangenome features based on pairwise, unsigned correlations. Labels indicate the number of OGC building methods (out of 4) that support each branch (values < 3 are not shown). In all cases when the support is not complete, reference database mapping is the method that disagrees. c Quantitative comparison of pangenome estimates among methods. Left: relative differences in pangenome estimates; right: relative contribution of methodological choices to between-species variance. Note the different color scale for Proteobacteria. Highlighted cells correspond to the features shown in d. d Species-wise comparison of selected pangenome features (pangenome size, number of single-copy core OGC, fluidity, and nucleotide sequence diversity in core genes) inferred from orthology- and synteny-based OGC. Black lines show the orthogonal least squares fit; gray lines indicate the 1:1 trend. Each point corresponds to the pangenome of one species, colored according to its phylum