Fig. 5. Analysis of strongly rhizosphere and growth-rate associated functions.

A A STRING search of the COGs that were most important for predicting rhizosphere-association in the RF models, and were enriched only in the rhizosphere and in fast growing MAGs in at least one phylum according to the PhyloGLM analysis. Top 200 most important COGs in the RF models were sorted decreasingly by feature importance, and only COGs that were associated with both rhizosphere and fast-growing MAGs and never with soil or slow growing MAGs in the PhyloGLM models were selected, resulting in 71 COGs uniquely associated with the rhizosphere and fast-growing MAGs. Edge weights represent the level of evidence for functional interaction according to STRING. Some relevant functions are colored according to legend. Common COGs among the top 50 important features in RF models to predict niche and growth rate status were labeled as “Top50 RS&CO models”. B Top 50 most important modules in DRAM-based RF models to predict niche and growth rate status were selected and only common features are represented in a barplot. Black circles represent modules that are also uniquely associated with rhizosphere and copiotrophs, while squares represent those only associated with oligotrophs and soil in the PhyloGLM results.