Fig. 3

Community assembly processes of NatComs and ecological association upon establishment of Pseudomonas protegens. Community assembly processes dominating the bacterial communities in response to the different community states, growing or stable A; different environments, bulk soil or wheat rhizosphere B; or the inoculation with P. protegens CHA0 C were calculated based on deviation from null community models. Differences in beta-nearest taxon index (βNTI), together with Raup-Crick-based Bray-Curtis served to determine the community assembly process that dominated the samples (bar plots). Per tested condition, all sampling times were merged. The threshold for the |βNTI| = 2 is highlighted as horizontal red lines. Differences in βNTI are based on Wilcoxon rank-sum test. Not significant (n.s.): p-value > 0.05. Ecological association inference is based on sparse inverse covariance estimation among the top 150 ASVs (dots, coloured according to the bacterial class assignation) from growing or stable community states, in bulk soil or the wheat rhizosphere in response to the inoculation with P. protegens CHA0 D or not inoculated E. Samples from the sampling times 5, 7 and 9 days were combined. Positive and negative interactions are coloured in red or green, respectively. Hub centrality scores of each ASV (dots) in the networks are based on the scaled Kleinberg’s hub centrality score. ASVs with a hub score ≥ 0.7 were considered keystone taxa and coloured according to the bacterial class assignation