Extended Data Fig. 7: Relationship between bacterial phylogenetic distance and niche preference difference (phylogenetic signal).

a, Mantel correlogram. b, Relationship curve and stepwise Mantel test. Mantel correlogram was performed as previously described. Stepwise Mantel test was used to evaluate the correlation between niche preference difference and phylogenetic distance within the phylogenetic distance from 0 to a certain value. On the relationship curves, each niche difference value is the mean value in each phylogenetic distance ‘class’ within a distance interval of 0.02. The niche preference difference was estimated based on nine representative environmental variables. In the nine variables, pH, U, and nitrate (NO₃) are major stressors in the contaminated area; Yb-sup, Sulfide, ⁵²Cr-ss, Cu-sup, Sr, and Rb-plt were identified as ‘centroid’ variables which showed the nearest distance to the centroids of six environmental variable clusters, to represent the variation of each cluster. The clusters were identified by hierarchical clustering based on pairwise correlation among environmental variables. The distances to centroid were calculated by multivariate homogeneity of groups dispersions using function ‘betadisper’ in R package ‘vegan’. P values (one-sided) are based on Mantel test (permutated 1000 times) and adjusted by false discovery rate (FDR) method. The Mantel correlograms and stepwise Mantel tests showed generally significant (P < 0.05) phylogenetic signal, validating the use of phylogenetic diversity to infer the influence of environmental selection. However, the trends in the relationship curves and the change of correlation coefficients and P values revealed the tipping points of phylogenetic signal at short phylogenetic distances around 0.2 to 0.6, supporting the necessity to use the phylogenetic-bin-based null model approach (iCAMP) which better exploits phylogenetic signal within relatively short phylogenetic distance.

Source data