Fig. 6

Runtime of different tools at different sample sizes. Runtime performance was evaluated on the simulated metagenomic datasets at different sample sizes with the same abundance of target strain spike-in. Note that the run time of cgMSI consists of three parts, namely, identifying candidate strains, MC sampling, and final result calling. Among them, the process of identifying candidate strains takes the most time. For ORI, we first mapped the samples to reference database using minimap2 to preliminarily filter the reads of the target species, and input these reads to ORI. Minimap2 outputs all secondary alignments by control parameter (–N 1000). For MetaMaps, we only counted the CPU time of the “map” stage due to the execution error of the “classify” stage