Methodology
Overall pipeline
The platform uses a four-stage pipeline: the TxGNN graph neural network predicts potential drug–disease associations, evidence is then collected automatically for each predicted pair, the evidence is graded from L1 to L5, and finally a decision recommendation is issued.
- TxGNN prediction: drug–disease relationships predicted by a graph neural network trained on a knowledge graph.
- Evidence collection: for each predicted pair, evidence is gathered from ClinicalTrials.gov, PubMed, DrugBank and Health Canada.
- Evidence grading: graded L1 to L5, where L1 is strongest (multiple Phase 3 RCTs) and L5 is model prediction only.
- Decision recommendation: Go, Proceed, Consider, Explore or Hold, based on the evidence level.
Evidence grading criteria
| Level | Definition | Clinical meaning |
|---|---|---|
| L1 | Multiple Phase 3 RCTs / systematic reviews | Strong support; clinical use may be considered |
| L2 | Single RCT or multiple Phase 2 trials | Moderate support; validation trials can be designed |
| L3 | Observational studies / large case series | Preliminary support; needs further validation |
| L4 | Preclinical / mechanistic studies | Theoretical support; far from clinical use |
| L5 | Model prediction only | Hypothesis stage; no human evidence yet |
How predictions are made
TxGNN’s deep learning (DL) model scores drug–disease pairs. Indications and contraindications already recorded for the drug in the TxGNN knowledge graph (KG) are then removed, so drug pages list only new predicted indications with a score of at least 0.99 (up to 10 per drug).
| Step | Role |
|---|---|
| Deep learning (DL) | The TxGNN graph neural network scores each drug–disease pair |
| Known-use check (KG) | Removes indications and contraindications already recorded in the knowledge graph; these are not counted as new predictions |
| Evidence collection | Searches ClinicalTrials.gov and PubMed for each remaining prediction and assigns an evidence level (L1–L5) |
Regulatory data integration
Drug approval data for Canada comes from Health Canada. Ingredient names are mapped to the DrugBank vocabulary; ingredients that cannot be mapped — herbal extracts, vaccines, excipients and others not catalogued by DrugBank — are excluded from prediction.
Limitations
- Predictions are statistical associations and do not imply causation or clinical efficacy.
- An L5 rating means model prediction only, with no supporting human evidence.
- Evidence collection depends on public databases; unpublished or unindexed studies are not captured.
- Ingredient mapping may miss items because of naming differences.
About the Developer
This platform is developed and operated by 藥提醒科技有限公司 (yao.care). See the About page for the full company and methodology background.
This report is for academic research reference only and does not constitute medical advice. Always follow your physician's instructions; never adjust medication on your own. Any drug repurposing decision requires full clinical validation and regulatory review.
Reviewed by: 藥提醒科技有限公司 (yao.care)