DrugBank Dataset
Dataset Overview
Our drug–drug interaction prediction model is trained on a curated dataset derived from DrugBank. The dataset has been carefully split into training, validation, and test sets to ensure robust model evaluation. The dataset is publicly available on Access Zenodo.
868,069
Total Drug Pairs
2,957
Unique Drugs
178
Interaction Types
3,780
Features SMILES
Download Dataset Splits
Access our preprocessed dataset splits for reproducible research
Citation
If you use this dataset in your research, please cite our work
Kha, Q.-H., Nguyen, D.-Q.-A., Pham, V.-H.-P., Huynh, K.-M.-U., Pham, D.-K., Phung, M.-T., Huynh, T.-P. et al. T-DDI: Robust Prediction of Drug Interactions using Chemical Descriptors. npj Digital Medicine (under review).