Skip to main content


Fig. 3 | Microbiome

Fig. 3

From: The evolutionary signal in metagenome phyletic profiles predicts many gene functions

Fig. 3

Gene co-evolution networks from metagenome PP and complete genome PP. a, c Nodes in the network represent COG gene families. Red nodes have assigned the GO function for which MPP-I was more predictive in cross-validation, blue nodes have the GO function where PP was more predictive and green nodes have assigned both functions. A network has two layers: red representing similarity (measured as Pearson correlation) between COG profiles in MPP-I data and blue in PP data. Green edges imply overlap across both layers. Similarities were computed using those metagenomes/genomes that had positive values of Random Forests feature importance (Gini method). b The part of MPP-I and PP data sets from which the network in panel a is constructed, showing COGs with two selected gene functions from a. Additionally, a random selection of negative control COGs that do not have these two functions annotated is shown. Rows are COGs, columns are selected features via Random Forest (complete genomes for PP, metagenomes for MPP), as in panel a. Both the rows and the columns were clustered using complete linkage hierarchical clustering method and Euclidean distance. MPP, metagenome phyletic profiles; PP, phyletic profiles

Back to article page