Table 1. Prediction Performance for the RF Models Using Different Featuresa.
molecular descriptor | Morgan FP | RDKit FP | MACCSkeys | |
---|---|---|---|---|
log loss | 0.561±0.012 | 0.592 ± 0.012 | 0.610 ± 0.009 | 0.592 ± 0.010 |
Acc | 0.722±0.016 | 0.684 ± 0.013 | 0.670 ± 0.026 | 0.687 ± 0.019 |
F1 score | 0.761±0.015 | 0.738 ± 0.020 | 0.724 ± 0.026 | 0.738 ± 0.018 |
AUC-ROC | 0.784±0.017 | 0.746 ± 0.017 | 0.724 ± 0.022 | 0.742 ± 0.012 |
The bold values represent the best scores. Prediction performance using the oral bioavailability test data set was reported in mean ± standard deviation.