Fig. 1. Replacement and modification rates vary across species, but only weakly depend on the diversity of the surrounding community.

Fraction of resident populations in which we detected strain replacement or evolutionary modification events (Methods), coarse-grained at the family (a) or phylum level (b). Lines denote 2.5–97.5 percentiles of the Gamma posterior distribution obtained from the observed number of counts (Methods). c The probability of a modification event as a function of the species diversity of the surrounding community. Points show the fraction of modification events in Bacteroidetes or Firmicutes species stratified by quantiles of the Shannon diversity; vertical lines denote 2.5–97.5 percentiles of the Gamma posterior distribution that were computed as above. Solid lines illustrate the best-fit logistic regression model using the phylum and community diversity as predictor variables (Methods), while the dashed lines show the average of each phylum computed for the bottom 90% of community diversity values. The analogous regression for strain replacements events was not statistically significant (P ≈ 0.8; Supplementary Fig. 3). d The distribution of the number of genetic changes per community. Bars denote the fraction of community comparisons with a given number of genetic events (replacements + modifications). These comparisons were further partitioned into two subgroups: those that only experienced modification events (left) and those that also experienced one or more replacement events (right). Lines denote the null distribution obtained by randomly permuting replacement and modification events across resident populations of the same species (Methods).