Skip to main content
Fig. 3 | BMC Bioinformatics

Fig. 3

From: Tree reconciliation combined with subsampling improves large scale inference of orthologous group hierarchies

Fig. 3

Domain benchmark results for tested OG definitions (S/SPLIT; M/MERGE; RND/RANDOM; Level Sampling/LS; 4.0/v4.0; 4.5/v4.5). (A’-A”) Example evaluation of InterPro entry IPR013241 (RNase P, subunit Pop3). F1 score and size (no. of proteins) are shown for the orthologous group (OG) with the best matching (F1) to annotated proteins (n=47), for each tested definition (colors) and selected taxonomic levels. Levels in nested order: Saccharomycetales, Ascomycota, Fungi, Opistokonta, Eukaryota. (B’-B”) Cumulative results for all tested domains (n=4120). For every OG definition (columns), for each domain, the best matching OG (F1 score) across all tested taxonomic levels is chosen. One-sided paired Wilcoxon signed rank test, alternative hypothesis F1(v4.0) - F1(LS) < 0, p-value: < 0.0001 (all cases); F1(v4.5) - F1(LS) < 0, p-value all domains: 0.06 (InterPro Family type 0.02 (Additional file 2: Figure S2A’) and non-significant for InterPro Domain type (Additional file 2: Figure S2A”)). (C) Selective comparison on 858 domains that differ more than 0.1 in F1 score between the compared methods (LS, v4.0, v4.5). Every point in the scatterplot represents F1 score and size of the best matching OG

Back to article page