User Guide

How to read and use EuTxGNN reports


Quick Start

1. Find a Drug of Interest

2. Check Evidence Level

Each drug-indication pair has an evidence level (L1-L5):

Level Meaning Confidence
L1 Multiple large RCTs support ⭐⭐⭐⭐⭐
L2 Single RCT or multiple Phase 2 ⭐⭐⭐⭐
L3 Observational studies ⭐⭐⭐
L4 Preclinical/Mechanistic research ⭐⭐
L5 AI prediction only

3. Review the Evidence

Each report contains detailed analysis and supporting data.


Understanding Predictions

Prediction Score

The TxGNN score (0-1) indicates model confidence:

Score Range Interpretation
0.99+ Very high confidence
0.90-0.99 High confidence
0.70-0.90 Moderate confidence
< 0.70 Lower confidence

Source Types

Source Description
TxGNN Knowledge Graph Based on network topology
TxGNN Deep Learning Based on neural network

FHIR Integration

For Developers

EuTxGNN provides FHIR R4 compliant APIs:

# Get metadata
curl https://eutxgnn.yao.care/fhir/metadata

# Get specific drug
curl https://eutxgnn.yao.care/fhir/MedicationKnowledge/metformin.json

For SMART Apps

Launch URL: https://eutxgnn.yao.care/smart/launch.html

Required scopes:

  • patient/MedicationRequest.read
  • patient/Patient.read

Data Export

CSV Format

Download predictions in CSV format from the Downloads page.

JSON Format

FHIR resources are available in JSON format via the API.


Limitations to Consider

  1. Research Only: Not for clinical decision-making
  2. Validation Required: All predictions need clinical validation
  3. Coverage: Limited to EMA-approved drugs with DrugBank mappings
  4. Evidence Levels: Currently all L5 (AI prediction only)

Frequently Asked Questions

What is drug repurposing?

Finding new therapeutic uses for existing approved drugs.

How accurate are the predictions?

TxGNN achieves high performance in validation studies, but predictions should be validated clinically.

Can I use this for clinical decisions?

No. This is for research purposes only.

How often is data updated?

Predictions are based on EMA data as of 2026. Updates may be made periodically.


Contact

For questions or feedback, please open an issue on GitHub.


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