Fig. 6

mOTUs2, Kraken, and Pathseq form a consenus with near-perfect genus-level classification performance. Box plots with individual data points for n = 11 simulated bacterial metagenomes showing genus-level F1 score (a), PPV (b), and SSV (c) for single tools, an intersection of classification between two tools, and a consensus of all three tools. PPV obtained perfect values in the result of an intersection between two tools or a consensus. Sensitivity generally decreases in the event of combining two tools with an intersection but increases to a median score of 0.905 in the result of an intersection. This raise in sensitivity resulted in a genus-level F1 score in the consensus approach of 0.95. mOTUs2 output files were unfiltered, whereas Kraken had a filter of >4 contigs and PathSeq >1 reads