Table 2.
Predictive performance of each feature groups by the 5-fold CV
| Feature groups | No. of features | Sn | Sp | Gm |
|---|---|---|---|---|
| Sequence features | 19 | 45.0 | 96.3 | 65.83 |
| Structure features | 30 | 82.8 | 97.6 | 89.90 |
| Base pair feature | 32 | 81.5 | 97.8 | 89.28 |
| Triplet sequence structure | 32 | 77.7 | 97.1 | 86.86 |
| SC related feature | 12 | 84.5 | 98.4 | 91.18 |
| All five feature groups | 125 | 84.0 | 98.3 | 90.86 |
| Feature SC | 1 | 76.6 | 96.9 | 86.15 |
| Feature SC × dP | 1 | 80.5 | 98.1 | 88.86 |
| Feature SC/(1 − dP) | 1 | 81.3 | 97.9 | 89.76 |
| Feature SCxdP/(1 − dP) | 1 | 82.2 | 98.2 | 89.84 |
| Feature SC × MFE/Mean_dG | 1 | 81.9 | 97.5 | 87.76 |
| Feature SC × zG | 1 | 78.9 | 98.8 | 89.36 |
| Feature SC/tot_bp | 1 | 0 | 100 | 0 |
| Feature SC/Len | 1 | 0 | 100 | 0 |
| Feature SC/NonBP_A | 1 | 68.5 | 98.4 | 82.09 |
| Feature SC/NonBP_C | 1 | 0 | 100 | 0 |
| Feature SC/NonBP_G | 1 | 0 | 100 | 0 |
| Feature SC/NonBP_U | 1 | 48.8 | 98.4 | 69.29 |
Sn = Sensitivity, Sp = Specificity and Gm = Geometric mean.