Figure 4.

Performance across pipelines for increment 2 in-silico-generated datasets. We report the F1-score obtained by each pipeline on the artificial datasets, using the SIB-provided database. Note that participants sometimes reported no viruses at all in some of the artificial samples, and we could therefore not calculate an F1-score for these samples (missing markers in the plot). In addition, as pipeline I had identified viruses in various settings (1 × 100 bp, 1 × 150 bp, 1 × 250 bp), we display here the results obtained only with 1 × 150 bp, which correspond to the sequencing settings they used in increment 1. Note that data points were jittered horizontally for readability.