SMG DDI Development Dataset

The scaffold data splitting method categorizes molecules according to their scaffold (molecular substructure). Priori studies  have shown that scaffold data splitting provides a more realistic estimate of model performance in prospective evaluation compared to random data splitting approach. Our dataset is based on three benchmark public datasets, ZhangDDI, ChCh-Miner, and DeepDDI. We apply scaffold data splitting method on the three datasets to create training, validation, and test data. The scaffold data splitting setting partitions the data based on two-dimensional structural frameworks, which include ring systems and linkers, providing a more realistic evaluation. The training, validation, and test datasets are used to validate the scalability and robustness of our proposed model SMG-DDI.