Extended Data Fig. 7: Predicting host taxonomy and environmental factors with random forest models.
From: Microbiomes of microscopic marine invertebrates do not reveal signatures of phylosymbiosis
Out-of-bag error rates of random forest models using microbial community ASVs to predict potential groupings. From left to right: all data, predicting community type (host-associated vs. environmental); animal-associated data, predicting host phylum, host phylum restricted to phyla that only include more than 20 specimens, host class, and host order; both animal-associated and environmental data, predicting location and habitat. The models can confidently discriminate microbial community type as well as environmental parameters from environmental communities. They fare poorly when discriminating any parameter from host-associated microbiomes, especially those related to host taxonomy.
