Fig. 4

Most significant associated bacterial taxa linked to high-halophyte, low-halophyte, and non-halophyte plants. (A) Microbial co-occurrence networks of the highest frequency (170). Nodes represent bacterial genera, and different colors indicate different clusters determined by the fast greedy algorithm. The keystone taxa were detected as a node above the 95% quantile of the fitted log-normal distribution of the four normalized network metrics (degree, betweenness, closeness centrality, and eigenvector centrality). (B) Top 50 important genera identified using the supervised machine learning algorithm (random forest). Statistical significance of selected bacterial genera was assessed by Kruskal-Wallis test