Drug Repurposing: From Data to Evidence

MyTxGNN is a drug repurposing prediction platform based on Harvard's TxGNN model. We predicted 41,560 potential new indications using knowledge graph methods, and 176,021 high-confidence predictions (score ≥ 0.7) using deep learning for 508 drugs approved by Malaysia's NPRA.

Leveraging AI to discover new therapeutic uses for existing medications. Our platform combines knowledge graphs with deep learning to identify promising drug-disease relationships for further clinical investigation.

Browse Drugs Learn Methodology


Key Statistics

508
Drugs Analyzed
41,560
KG Predictions
176,021
High-Score DL Predictions
27,938
NPRA Registered Products

Our Approach

MyTxGNN uses two complementary approaches for drug repurposing prediction: Knowledge Graph (KG) methods leverage existing drug-disease relationships, while Deep Learning (DL) models learn complex patterns from biomedical data.

Knowledge Graph Predictions
Based on TxGNN's biomedical knowledge graph containing drug-disease relationships from DrugBank, clinical trials, and scientific literature. 41,560 predictions for 508 drugs.
Deep Learning Predictions
TxGNN's neural network model provides confidence scores for each drug-disease pair. 176,021 predictions with score ≥ 0.7, indicating high confidence.
Malaysia NPRA Integration
Focused on drugs registered with Malaysia's National Pharmaceutical Regulatory Agency (NPRA), ensuring relevance to local healthcare needs.
Evidence Collection
Automated collection of supporting evidence from ClinicalTrials.gov, PubMed, and other authoritative sources to validate predictions.

Top Predictions by Disease Category

Disease Category Predictions Top Drug
Allergic Rhinitis 392 Prednisolone, Betamethasone
Hypertension 366 Multiple antihypertensives
Rheumatoid Arthritis 316 Corticosteroids
Seborrheic Dermatitis 288 Ketoconazole, Fusidic Acid
Asthma 259 Bronchodilators, Steroids

Quick Navigation

Section Description Link
Drug List Browse all 508 analyzed drugs View Drugs
Methodology Learn about our prediction approach Read More
Data Sources Explore our data sources View Sources
About Learn about this project About Us
Downloads Get prediction data Download
FHIR API Access via FHIR R4 API Docs

About This Project

This platform uses the TxGNN deep learning model published in Nature Medicine by Harvard University's Zitnik Lab to predict potential new indications for Malaysia NPRA-approved drugs.

"TxGNN is the first foundation model for drug repurposing designed specifically for clinicians, integrating knowledge graphs with deep learning to predict drug efficacy for rare diseases." — Huang et al., Nature Medicine (2023)

Data Scale

Item Count
NPRA Registered Products 27,938
Drugs with DrugBank Mapping 508
KG Predictions 41,560
DL High-Score Predictions 176,021
Unique Diseases 17,041

Data Sources

This platform integrates multiple authoritative public data sources to ensure prediction traceability and academic value.


Disclaimer
This report is for academic research purposes only and does not constitute medical advice. Drug use should follow physician guidance. Do not self-adjust medications. Any drug repurposing decisions require complete clinical validation and regulatory review.

Last updated: 2026-03-03 | MyTxGNN Research Team

Copyright © 2026 MyTxGNN Project. For research purposes only. Not medical advice.