Table 4 Performance comparison of Acc, AUC, and Prop on closely and distantly related datasets
From: PCVR: a pre-trained contextualized visual representation for DNA sequence classification
Method | Closely related | Distantly related | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Â | Supk. | Phyl. | Supk. | Phyl. | ||||||||
| Â | Acc | AUC | Prop | Acc | AUC | Prop | Acc | AUC | Prop | Acc | AUC | Prop |
MMseqs2 | 99.63 | 0.94 | 87.45 | 97.30 | 0.93 | 87.45 | 90.73 | 0.75 | 52.78 | 75.04 | 0.71 | 52.78 |
MMseqs2 tax. | 98.06 | 0.96 | 92.62 | 92.62 | 0.93 | 92.69 | 80.50 | 0.79 | 71.37 | 59.19 | 0.73 | 71.37 |
minimap2 | 99.95 | 0.90 | 75.59 | 99.55 | 0.88 | 75.59 | 94.42 | 0.62 | 19.88 | 77.12 | 0.59 | 19.88 |
DeepMicrobes | 96.68 | 0.98 | 100 | 87.72 | 0.94 | 100 | 68.39 | 0.81 | 100 | 41.95 | 0.70 | 100 |
BERTax | 94.78 | 0.97 | 100 | 85.55 | 0.93 | 100 | 88.95 | 0.94 | 100 | 60.10 | 0.80 | 100 |
PCVR-Base | 98.57 | 0.99 | 100 | 95.08 | 0.99 | 100 | 95.31 | 0.99 | 100 | 69.38 | 0.95 | 100 |
PCVR-Large | 98.99 | 0.99 | 100 | 96.09 | 0.99 | 100 | 96.48 | 0.98 | 100 | 74.69 | 0.92 | 100 |