{"architecture":{"family":"Graph neural network with disease-similarity metric learning","input":"knowledge graph (drugs, diseases, genes, phenotypes)","output":"indication / contraindication scores with explanatory paths"},"article":null,"cancer_slugs":["pan-cancer"],"category":"drug-discovery","confidence":"medium","datasets":[],"developer":"Zitnik Lab, Harvard Medical School","evaluation":[{"benchmark":"zero-shot held-out diseases","external":true,"metric":"AUPRC improvement over baselines","source":"https://www.nature.com/articles/s41591-024-03233-x","value":"49.2% (indications), 35.1% (contraindications), per paper"}],"hf":null,"kind":"task-model","license":"MIT","limitations":"- Predictions are hypotheses over knowledge-graph edges; clinical evidence is required before any use.\n- Oncology indications are a subset of a general-purpose graph.","links":{"demo":null,"docs":null,"doi":"10.1038/s41591-024-03233-x","github":"https://github.com/mims-harvard/TxGNN","huggingface":null,"paper":"https://www.nature.com/articles/s41591-024-03233-x","pmid":null},"modalities":["molecules","literature"],"name":"TxGNN","notable_uses":"","openness":"open-weights","regulatory":{"intended_use_en":"Hypothesis generation for repurposing research.","intended_use_pl":"Generowanie hipotez do bada\u0144 nad repozycjonowaniem.","source_url":"https://github.com/mims-harvard/TxGNN","status":"research-only"},"regulatory_status":"research-only","release_date":"2024-09-25","run_snippet":"","settings":["drug-discovery"],"slug":"txgnn","sources":[{"label":"Huang K et al. A foundation model for clinician-centered drug repurposing. Nat Med 2024","url":"https://www.nature.com/articles/s41591-024-03233-x"}],"summary":"Graph neural network for zero-shot drug repurposing that scores drug\u2013disease indications and contraindications over a medical knowledge graph of 17,080 diseases, including ones with no approved treatment.","tasks":["drug-repurposing","drug-discovery"],"training":{"summary_en":"PrimeKG-derived medical knowledge graph; evaluated on diseases held out entirely from training (zero-shot).","summary_pl":"Graf wiedzy medycznej pochodny od PrimeKG; oceniany na chorobach ca\u0142kowicie wy\u0142\u0105czonych z treningu (zero-shot)."},"updated_at":"2026-09-05T22:26:01.079119","url":"/ai-oncology/models/txgnn","usage":{"library":"PyTorch Geometric (repo)","notes_en":"Pretrained model and an explorer UI are linked from the repository."},"verified_at":"2026-09-05T22:26:01.078860","verified_by":"editorial","version":null,"what_it_does":""}