JpTxGNN: Drug Repurposing Predictions for Japanese Medicines

Overview

JpTxGNN is a drug repurposing prediction platform for Japanese medicines, based on Harvard's TxGNN deep learning model. The system predicts potential new indications for approved medications using knowledge graph and deep learning approaches.

Key Features

Dual Prediction Methods

Method	Description	Speed	Accuracy
Knowledge Graph (KG)	Query existing drug-disease relationships in TxGNN knowledge graph	Fast (minutes)	Medium
Deep Learning (DL)	Neural network model prediction with confidence scores	Slow (hours)	High

Japanese Medicine Focus

SSK Medical Drugs: 19,317 prescription medicines from Japan
KEGG DRUG Integration: Therapeutic classification and indication information
DrugBank Mapping: International drug identifier standardization

FHIR R4 Compliant API

MedicationKnowledge: Drug information resources
ClinicalUseDefinition: Predicted indication resources
Bundle: Collection of all predictions

Statistics

Metric	Value
Total Drugs	3,824
DrugBank Mappings	142
KG Predictions	33,901
DL Predictions	2,419,822
Integrated Predictions (≥90%)	155,638

Prediction Workflow

Data Collection: Integrate SSK medicines + KEGG therapeutic information
DrugBank Mapping: Map ingredient names to DrugBank IDs
KG Prediction: Extract known relationships from TxGNN knowledge graph
DL Prediction: Predict new relationships using deep learning model
Integration & Filtering: Extract predictions with confidence ≥90%

TxGNN Score Interpretation

The TxGNN score represents model confidence for drug-disease pairs, ranging from 0 to 1.

Threshold	Meaning	Recommended Use
≥ 0.99	Very high confidence	Priority verification
≥ 0.90	High confidence	Detailed investigation
≥ 0.50	Medium confidence	Reference information
< 0.50	Low confidence	Additional validation needed

Technical Stack

Backend: Python, pandas, PyTorch, DGL
Frontend: Jekyll, JavaScript, Fuse.js
API: HL7 FHIR R4
Hosting: GitHub Pages

Data Sources

Data	Source	Description
Medicines	Japan SSK	19,317 prescription medicines
Therapeutic Info	KEGG DRUG	Indications and effects
Knowledge Graph	Harvard TxGNN	17,080 diseases, 7,957 drugs

Disclaimer

This project is for research purposes only and does not constitute medical advice. Drug repurposing candidates require clinical validation before application.

Citation

If you use this dataset or software, please cite:

@software{jptxgnn2026,
  author       = {Yao.Care},
  title        = {JpTxGNN: Drug Repurposing Predictions for Japanese Medicines},
  year         = 2026,
  url          = {https://github.com/yao-care/JpTxGNN}
}

Also cite the original TxGNN paper:

@article{huang2024txgnn,
  title={A foundation model for clinician-centered drug repurposing},
  author={Huang, Kexin and Chandak, Payal and Wang, Qianwen and Haber, Shreyas and Zitnik, Marinka},
  journal={Nature Medicine},
  year={2024},
  doi={10.1038/s41591-024-03233-x}
}