Model performance from 5-fold CV with scaffold-splitting
| SVM | LR | XGB | RF | DNN | |
|---|---|---|---|---|---|
| Accuracy | 0.906 ± 0.005 | 0.905 ± 0.007 | 0.906 ± 0.007 | 0.904 ± 0.006 | 0.801 ± 0.03 |
| Precision | 0.902 ± 0.005 | 0.918 ± 0.007 | 0.917 ± 0.007 | 0.9 ± 0.006 | 0.834 ± 0.041 |
| Recall | 0.958 ± 0.005 | 0.935 ± 0.006 | 0.939 ± 0.008 | 0.958 ± 0.006 | 0.811 ± 0.064 |
| F1 score | 0.929 ± 0.004 | 0.926 ± 0.005 | 0.927 ± 0.006 | 0.927 ± 0.004 | 0.818 ± 0.036 |
| AUROC | 0.968 ± 0.003 | 0.965 ± 0.003 | 0.964 ± 0.004 | 0.968 ± 0.003 | 0.886 ± 0.028 |
Performance metrics were obtained across 100 iterations of 5-fold CV with scaffold-splitting (average ± standard deviation). CV, cross-validation; SVM, Support Vector Machine; LR, Logistic Regression; XGB, XGBoost; RF, Random Forest; DNN, Deep Neural Network, AUROC, area under the receiver operating characteristic curve.