Fig. 2

Performance estimates for taxonomic classification tools. Methods were applied to quality filtered and human-depleted sequencing reads on 100 metagenome simulations. Performance is summarized at the genus level in terms of sensitivity (a), positive predictive value (b), and F1 score (c). Computational resources in terms of CPU time and RAM is also shown for the top two performing tools: Kraken and mOTUs2 (d). Kraken utilized 20 threads for most datasets whereas mOTUs2 utilized 17. mOTUs2 output was unfiltered, whereas Kraken had a confidence threshold of 0.2 and a subsequent read threshold of 500 applied to determine positive classifications. Parameters for each tool in this graphic were selected from the top-performing parameters observed for multiple tests with varying parameters