Table 1 Benchmarking of taxonomic assignment tasks using simulated reads (ICTV and HIV-1)
From: A multi-task CNN learning model for taxonomic assignment of human viruses
Datasets | Simulated reads (50 bp, n = 1million) from ICTV standard genomes | Simulated reads (150 bp, n = 100 k) from ICTV standard genomes | Simulated divergent HIV-1 reads (150 bp, n = 100 k) (100 generations) | Simulated divergent HIV-1 reads (150 bp, n = 100 k) (1 K generations) | Simulated divergent HIV-1 reads (150 bp, n = 100 k) (10 K generations) | ||
|---|---|---|---|---|---|---|---|
Metrics | Kappa | F1 | Kappa | F1 | Prediction accuracy | Prediction accuracy | Prediction accuracy |
Kraken 2 | 0.3139 | 0.3184 | 0.7274 | 0.7318 | 0.9664 | 0.7755 | 4.890 × 10–3 |
Centrifuge | 0.5953 | 0.6011 | 0.7219 | 0.7266 | 0.8865 | 0.8679 | 3.056 × 10–2 |
Bowtie 2 | 0.9597 | 0.9605 | 0.9991 | 0.9991 | 0.9976 | 0.9828 | 8.501 × 10–4 |
MT-CNN | 0.9464 | 0.9475 | 0.9976 | 0.9976 | 1.000 | 0.9963 | 0.1531 |