AbdulrahmanMahmoud007/DDI-KG-SS-GNN

DDI KG-SS-GNN Streamlit Demo

This Hugging Face Space hosts the Streamlit inference demo for drug-drug interaction prediction using the KG-SS-GNN model.

The app includes:

• —selection between the cold_split and semi_cold_split checkpoints
• —prediction of the top interaction classes for two SMILES strings
• —molecule rendering, graph statistics, BRICS fragments, fingerprint density, KG-proxy descriptor views, and prediction confidence

Files

app.py
requirements.txt
src/ddi_project/
checkpoint/
  kgssgnn_best_pickle_cold_split.pkl
  kgssgnn_best_pickle_semi_cold_split.pkl
data/
  unseen_demo_pairs.csv

The full DrugBank training CSV is not required for the live demo. The demo uses the trained checkpoints and accepts either preset or custom SMILES inputs.