Fig. 4From: Inference-based accuracy of metagenome prediction tools varies across sample types and functional categoriesThe results of inference methods in unpermuted and permuted datasets. The red points are the inference correlations between metagenome prediction tools and unpermuted metagenome sequencing data for each dataset. The boxplots of blue points show the inference correlations between metagenome prediction tools and permuted metagenome sequencing data for 100 permutations. In each of the 100 permutations, every gene’s abundance was permuted across samples independentlyBack to article page