Performance of three machine learning methods (k-NN, PNN, and SVM) for predicting PXR and hPXR activators and nonactivators determined by a 10-fold cross-validation study
The results are expressed in SE (sensitivity or prediction accuracy for PXR activators), SP (specificity or prediction accuracy for PXR non-activators), Q (overall accuracy), and C (Matthews' correlation coefficient).
Molecular Descriptors and Method | SE | SP | Q | C |
|---|---|---|---|---|
| % | % | % | ||
| All Species | ||||
| All descriptors | ||||
| k-NN | 81.7 | 57.5 | 72.6 | 0.410 |
| PNN | 81.9 | 60.9 | 74.0 | 0.446 |
| SVM | 81.0 | 62.2 | 73.9 | 0.441 |
| RFE-selected descriptors | ||||
| k-NN | 84.0 | 61.2 | 75.4 | 0.473 |
| PNN | 82.8 | 68.4 | 77.4 | 0.526 |
| SVM | 81.2 | 70.3 | 77.1 | 0.528 |
| Human | ||||
| All descriptors | ||||
| k-NN | 80.6 | 62.4 | 72.5 | 0.448 |
| PNN | 80.7 | 63.8 | 73.2 | 0.461 |
| SVM | 77.8 | 71.4 | 74.9 | 0.504 |
| RFE-selected descriptors | ||||
| k-NN | 80.8 | 67.7 | 75.0 | 0.499 |
| PNN | 85.0 | 68.7 | 77.7 | 0.559 |
| SVM | 84.4 | 73.6 | 79.6 | 0.598 |