Fig. 3

OTUs responsible for the differences during domestication of rice are revealed by a differential abundance test and random forest classification. a, b Volcano plot visualizing differentially abundant a bacterial and b fungal OTUs between the 17 wild and 26 domesticated rice accessions. The comparison was made using a zero-inflated Gaussian distribution mixture model on the CSS normalized OTU tables followed by a moderated t test and a Bayesian approach. Data from all three replicates of each accession were used. This test revealed 39 bacterial and 18 fungal OTUs that were significantly enriched in domesticated rice (above four folds (log2 Fold change > 2), FDR < 0.01). A total of 224 bacterial and 167 fungal OTUs were enriched in wild rice (above 4 folds (log2 Fold change < − 2), FDR < 0.01). c, d Relative abundance profiles for the Top 20 operational taxonomic units (OTUs) best discriminating wild and domesticated rice accessions with Random forest (RF) classifier in c bacterial and d fungal communities. OTUs are colored by their categorization as ‘wild-enriched,’ ‘domesticated-enriched,’ and ‘non-differential’ according to the result of a differential abundance test in a and b. The RF model with top 20 OTUs were constructed with a 10-fold cross validation method. OTUs were ranked by their importance in contributing to the accuracy of wild and domesticated rice prediction in the RF model by calculating the mean decrease in Gini coefficient. OTUs are ordered along the y-axis by rank of importance within each of the three categories