Table 7.
Prediction results of various hERG blockade classification models developed with training sets different from Doddareddy's training set.
| Entry | Model | Training set | Test set | SE | SP | Q |
|---|---|---|---|---|---|---|
| 1 | RP (Wang et al., 2016) | Hou's training set 1 (P: 283; N: 109) | Hou's test set 1 (P: 129; N: 66) | 79.8% | 75.8% | 78.5% |
| NB (Wang et al., 2016) | 82.2% | 75.8% | 80.0% | |||
| SVM (Wang et al., 2016) | 90.7% | 65.2% | 82.1% | |||
| Conv-CapsNet | 85.7% | 78.8% | 82.0% | |||
| RBM-CapsNet | 84.1% | 80.3% | 82.0% | |||
| 2 | RP (Wang et al., 2016) | Hou's training set 2 (P: 272; N: 120) | Hou's test set 2 (P: 140; N: 55) | 80.0% | 74.5% | 78.5% |
| NB (Wang et al., 2016) | 81.4% | 80.0% | 81.0% | |||
| SVM (Wang et al., 2016) | 85.0% | 74.5% | 82.1% | |||
| Conv-CapsNet | 82.1% | 81.8% | 82.0% | |||
| RBM-CapsNet | 81.4% | 83.6% | 82.0% | |||
| 3 | Bayesian (Wang et al., 2012) | Hou's training set 3 (P: 314; N: 306) | Hou's test set 3 (P: 63; N: 57) | 86.9% | 83.1% | 85.0% |
| Conv-CapsNet | 87.3% | 86.0% | 86.8% | |||
| RBM-CapsNet | 88.9% | 84.2% | 86.8% | |||
| 4 | SVM (Zhang et al., 2016) | Zhang's training set (P: 717; N: 210) | Zhang's test set (P: 188; N: 48) | 95.8% | 34.0% | 83.5% |
| kNN (Zhang et al., 2016) | 92.6% | 40.4% | 82.2% | |||
| Conv-CapsNet | 88.8% | 66.7% | 84.5% | |||
| RBM-CapsNet | 90.4% | 64.6% | 85.2% | |||
| 5 | LibSVM (Siramshetty et al., 2018) | Sun's training set (P: 483; N: 2541) | Sun's test set (P: 53; N: 13) | 68.0% | 85.0% | 71.0% |
| RF (Siramshetty et al., 2018) | 72.0% | 85.0% | 74.0% | |||
| Conv-CapsNet | 83.0% | 84.6% | 83.3% | |||
| RBM-CapsNet | 86.8% | 84.6% | 86.3% | |||
| 6 | LibSVM (Siramshetty et al., 2018) | Siramshetty's training set T3 (P: 1406; N: 1708) | Doddareddy's test set (P: 108; N: 147) | 64.0% | 89.0% | 78.0% |
| RF (Siramshetty et al., 2018) | 68.0% | 91.0% | 81.0% | |||
| Conv-CapsNet | 85.2% | 88.4% | 87.1% | |||
| RBM-CapsNet | 83.3% | 91.2% | 87.8% |