AI Tool Identifies Drug Candidates for Over 17,000 Rare Diseases

Harvard's TxGNN model repurposes existing medicines, offering new treatment possibilities for conditions with limited or no options.

TxGNN, developed by Harvard Medical School, is the first AI model specifically designed to identify drug candidates for rare and untreated diseases.
The tool analyzed nearly 8,000 medicines and identified potential treatments for 17,080 diseases, significantly more than any previous AI model.
TxGNN's predictions include potential side effects and contraindications, providing a safer and more effective approach to drug repurposing.
The model's zero-shot inference capability allows it to suggest treatments for diseases it has never encountered before, enhancing its versatility.
TxGNN is available for free to clinician-scientists, aiming to bridge the treatment gap for rare, ultrarare, and neglected conditions.