MVA-DDI

MVA-DDI: Interpretable attention network with multi-view learning for drug-drug interaction prediction

File description

• Data: Pre-split training set, test set, and validation set.

• Encoder/deal_data.py: SMILES data preprocessing.

• ESPF: Encoding dictionary and corresponding fields.

• Model/MVA.py: Model code.

• log.py: Evaluation metrics calculation.

• losses.py: Implementation of cross-entropy loss function.

• Result_vis.py: Visualization of loss and roc results.

• train.py: Main function for model training and testing.

• violin.py: Drawing violin analysis plot.

Requirement

• Python == 3.7

• Pytorch

• RDKit

• scikit-learn

• subword_nmt

Usage

python train.py

Dataset

Our dataset comes from DrugBank (V5.1.9) and ChEMBL (V32). We provide the preprocessed dataset, but you can also directly download the original DrugBank dataset V5.1.9 and ChEMBL dataset V32