Drug-Drug Interaction prediction on DDI dataset (5-fold cross-val)
95.7AUCEnsemble method (L1)
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Ensemble method (L1)features=substructure, target, enzyme, transporter, pathway, indication, and off side effect, ensemble_type=L1, source=cited from Zhang et al., 20172021.09 | 95.7 | 80.7 | 78.5 | 67 | |
| Ensemble method (L2)features=substructure, target, enzyme, transporter, pathway, indication, and off side effect, ensemble_type=L2, source=cited from Zhang et al., 20172021.09 | 95.6 | 80.6 | 78.3 | 66.5 | |
| DPDDIarchitecture=GNN, input=molecule graph2021.09 | 95.6 | 90.7 | 75.4 | 81 | |
| MM-Deacon (concat)input_modality=SMILES + IUPAC, embeddings=pre-trained models, classifier=MLP2021.09 | 95 | 91.8 | 81.9 | 82.4 | |
| MM-Deacon (IUPAC)input_modality=IUPAC, embeddings=pre-trained models, classifier=MLP2021.09 | 94.7 | 91.3 | 83.4 | 79.7 | |
| MM-Deacon (SMILES)input_modality=SMILES, embeddings=pre-trained models, classifier=MLP2021.09 | 94.6 | 91.1 | 80.5 | 82.3 | |
| MLM-[CLS]embeddings=pre-trained models, classifier=MLP2021.09 | 94.3 | 90.1 | 78.4 | 81.3 | |
| Neighbor recommenderdrug structural features=true, source=cited from Zhang et al., 20172021.09 | 93.6 | 75.9 | 76.5 | 61.7 | |
| Random walkdrug structural features=true, source=cited from Zhang et al., 20172021.09 | 93.6 | 75.8 | 76.3 | 61.6 |