Fig. 3.

Comparison of precision, recall and F1 score across reference databases and classifiers. Genomic DNA was extracted in duplicate with the Zymo Research Quick-DNA Fecal/Soil Microbe Miniprep Kit. The number of reads classified as Listeria monocytogenes was averaged across reference databases and classifiers for each set of replicates. Precision, recall and F1-score were calculated for samples artificially contaminated with L. monocytogenes colony forming units (CFU) at three levels (high:10⁴ CFU/ml; medium 10³ CFU/ml; low: 10² CFU/ml). Raw = raw datasets classified with default settings (no host sequence removal); Filtered = low-quality read filtered datasets (fastp) classified with default settings; Filtered + HR = low-quality read filtered (fastp) and in silico host removed (Bowtie 2) datasets classified with default settings; UNIVERSAL = universal (taxon-agnostic) reference database; FBP = foodborne pathogen-specific (taxon-specific) reference database. *No L. monocytogenes reads were detected in RTE turkey deli meat low spike-in datasets