Fig. 3

Potential keystone taxa for CH₄ emissions. a Contribution of five groups of methanogenic communities to CH₄ emissions based on random forest regression model. Five groups were classified based on the frequency of co-occurrence relationships between pairwise OTUs (A.endemic, always endemic group; C.endemic, conditionally endemic group; Moderate, moderate group; C.common, conditionally common group; A.common, always common group). b Biomarker taxa for CH₄ emissions. The top 19 genera were identified by applying random forest classification of the relative abundance. Biomarker taxa are ranked in descending order of importance to the accuracy of the model. The inset represents tenfold cross-validation error as a function of the number of input genera used to differentiate microbiota in order of variable importance. Among the 19 genera, 7 genera contain the keystone OTUs (33 OTUs involved in C.common and A.common) and account for 47% of total microbial production of CH₄ emissions; the other 12 genera account for 53%. c Functional co-occurrence network of 33 keystone OTUs and functional genes involved in CH₄ generation. d Linear regression analysis of the connections between genes and species with CH₄ emissions. Blue shaded area indicates 95% confidence interval (Spearman’s P < 0.05)