Table 4. Performance of binary profile-based SVM models for predicting hemotoxicity of peptides using five-fold cross-validation.
| Approach | Sn (%) | Sp (%) | Acc (%) | MCC |
|---|---|---|---|---|
| HemoPI-1 main dataset | ||||
| NT10 | 90.5 | 96.1 | 93.7 | 0.87 |
| CT10 | 85.0 | 93.4 | 89.6 | 0.79 |
| NTCT10 | 93.8 | 97.5 | 95.7 | 0.91 |
| NT15 | 92.7 | 97.0 | 95.0 | 0.90 |
| CT15 | 89.0 | 91.5 | 90.3 | 0.81 |
| NTCT15 | 94.0 | 96.7 | 95.3 | 0.91 |
| HemoPI-2 main dataset | ||||
| NT10 | 70.7 | 72.2 | 71.4 | 0.43 |
| CT10 | 70.0 | 71.4 | 70.6 | 0.41 |
| NTCT10 | 78.9 | 74.6 | 77.1 | 0.54 |
| NT15 | 70.9 | 75.8 | 73.1 | 0.46 |
| CT15 | 67.5 | 77.4 | 71.8 | 0.45 |
| NTCT15 | 77.0 | 77.0 | 77.0 | 0.54 |
Sn: Sensitivity; Sp: Specificity; Acc: Accuracy; MCC: Matthews correlation coefficient; NT10: first 10 residues from N-terminus; CT10: last 10 residues from C-terminus; NTCT10: NT10+CT10.