Table 9 The top three model and corresponding performances based on the results from 50 times independent runs for each dataset. (the descriptor-based models were colored as italic and the graph-based model were colored as undeline)
Dataset | No. | Tasks | Metric | Top 1 | Top 2 | Top 3 |
|---|---|---|---|---|---|---|
ESOL | 1127 | 1 | RMSE | SVM (0.569 ± 0.052) | XGBoost (0.582 ± 0.056) | Attentive FP (0.587 ± 0.065) |
FreeSolv | 639 | 1 | RMSE | SVM (0.852 ± 0.171) | DNN (1.013 ± 0.197) | XGBoost (1.025 ± 0.185) |
Lipop | 4200 | 1 | RMSE | Attentive FP (0.553 ± 0.035) | XGBoost (0.574 ± 0.034) | SVM (0.577 ± 0.039) |
HIV | 40748 | 1 | AUC_ROC | GCN (0.834 ± 0.025) | GAT (0.826 ± 0.030) | SVM (0.822 ± 0.020) |
BACE | 1513 | 1 | AUC_ROC | GCN (0.898 ± 0.019) | SVM (0.893 ± 0.020) | RF (0.890 ± 0.022) |
BBBP | 2035 | 1 | AUC_ROC | RF (0.927 ± 0.025) | XGBoost (0.926 ± 0.026) | DNN (0.922 ± 0.029) |
ClinTox | 1475 | 2 | AUC_ROC | XGBoost (0.911 ± 0.036) | RF (0.911 ± 0.042) | Attentive FP (0.904 ± 0.043) |
SIDER | 1366 | 27 | AUC_ROC | RF (0.646 ± 0.022) | XGBoost (0.642 ± 0.020) | GCN (0.634 ± 0.026) |
Tox21 | 7811 | 12 | AUC_ROC | Attentive FP (0.852 ± 0.012) | DNN (0.840 ± 0.014) | RF (0.838 ± 0.011) |
ToxCast | 8539 | 182 | AUC_ROC | Attentive FP (0.794 ± 0.017) | DNN (0.786 ± 0.019) | RF (0.782 ± 0.005) |
MUV | 93087 | 17 | AUC_PRC | SVM (0.112 ± 0.045) | XGBoost (0.068 ± 0.028) | RF (0.061 ± 0.032) |