Table 8 A summary of predictive scores for the BBB dataset as presented in the work by [2] followed by the performance of the InChI-NGN. Missing ranges for all values other than the SVM work and our own are due to missing information in the primary literature.
Method | Q(%) | SE(%) | SP(%) | MCC |
|---|---|---|---|---|
LR | 57.1 | 63.6 | 42.8 | 0.063 |
LDA | 46.8 | 40.0 | 58.4 | −0.067 |
C4.5 DT | 73.8 | 83.7 | 54.9 | 0.398 |
k-NN | 70.8 | 77.0 | 58.0 | 0.348 |
PNN | 70.3 | 76.2 | 57.8 | 0.357 |
SVM | 71.0±4.53 | 89.9±3.16 | 64.3±13.07 | 0.524±0.117 |
LR RFE | 71.0 | 83.9 | 46.4 | 0.321 |
LDA RFE | 71.2 | 78.2 | 58.3 | 0.360 |
C4.5 DT RFE | 74.3 | 80.3 | 62.8 | 0.433 |
k-NN RFE | 77.1 | 85.5 | 61.4 | 0.477 |
PNN RFE | 76.1 | 84.3 | 62.1 | 0.481 |
SVM RFE | 83.7±3.90 | 88.6±7.01 | 75.0±12.83 | 0.645±0.080 |
InChI-NGN | 72.0±2.33 | 77.6±1.72 | 59.0±3.91 | 0.355±0.052 |