Table 1.
Performance of the best models on the external test set
Dataset | Highest balanced accuracy | Highest AUC | ||
---|---|---|---|---|
Balanced accuracy | AUC | Balanced accuracy | AUC | |
1-1 | 0.664 | 0.811 | 0.610 | 0.814 |
1-16 | 0.760 | 0.805 | 0.678 | 0.818 |
2-2 | 0.680 | 0.801 | 0.561 | 0.821 |
2-32 | 0.753 | 0.815 | 0.753 | 0.815 |
5-5 | 0.630 | 0.814 | 0.551 | 0.823 |
5-80 | 0.784 | 0.803 | 0.693 | 0.823 |
SMOTE | 0.740 | 0.813 | 0.740 | 0.813 |
Tox21 highest | 0.765 | 0.847 | 0.581 | 0.880 |
The performance for the best single models trained with COVER and SMOTE as well as the best models for AUC and balanced accuracy from the Tox21 challenge.The best models are highlighted with italics numbers