MUDI: A Multimodal Biomedical Dataset for Understanding Pharmacodynamic Drug-Drug Interactions

This repository provides the full dataset, molecular representations, textual metadata, and preprocessing scripts used in the MUDI project. The dataset is designed for developing and benchmarking multimodal models that predict clinically meaningful drug–drug interactions (DDIs) from a pharmacodynamic perspective. 📁 Directory Structure . ├── dataset/ │ ├── train.csv │ └── test.csv │ ├── molecules/ │ ├── images/ │ │ └── [drug_id].png │ └── graphs/ │ └── [drug_id].graphml │ └── drug_info.json 📄 Dataset Overview dataset/train.csv and dataset/test.csv These two CSV files contain the labeled DDI pairs used for training and evaluation. Each file has the following columns: Drug1: Unique identifier for the first drug Interaction: One of the pharmacodynamic classes: Synergism: Drug1 enhances the effect of Drug2 Antagonism: Drug1 reduces or neutralizes the effect of Drug2 New Effect (New Adverse): The combination causes a novel effect not present when the drugs are used separately Drug2: Unique identifier for the second drug All drug IDs correspond to keys in drug_info.json. 🧪 Molecular Representations molecules/images/ Contains 2D structure diagrams of each drug, rendered from its SMILES representation. File format: .png Naming convention: [drug_id].png Resolution: 1000×800 pixels Can be used with any standard image model (e.g., Vision Transformer) molecules/graphs/ Contains molecular graphs derived from SMILES, encoded in GraphML format. File format: .graphml Naming convention: [drug_id].graphml Standard: Follows the GraphML specification Recommended library: NetworkX Structure: Nodes: Atoms Edges: Bonds between atoms (single, double, aromatic, etc.) 📚 Textual and Structural Metadata drug_info.json Contains comprehensive textual and structural information for every drug used in the dataset. Each entry has the following schema: { "drug_id": { "name": "DrugName", "description": { "summary": "...", "indication": "...", "metabolism": "...", "moa": "...", "pharmacodynamics": "..." }, "formula": "C20H25N3O", "smiles": "CC(C)NCC(O)..." } } These fields are used to build input features for textual and formula-based encoders. 📜 License and Citation License: Creative Commons Attribution-NonCommercial 4.0 (CC BY-NC 4.0) If you use this dataset in your research, please cite: to be announced 🤝 Contact For questions, feedback, or collaboration inquiries, please contact: 📧 [[e-mail]]