Understanding Evidence Levels
SgTxGNN uses a five-level evidence classification system (L1-L5) to help researchers quickly assess the credibility of drug repurposing predictions.
The Five Levels
L1 - Multiple Phase 3 RCTs
The highest level of evidence. Predictions at this level are supported by:
- Two or more completed Phase 3 randomized controlled trials
- Consistent positive results across trials
- Large patient populations
Recommendation: Strong candidates for clinical evaluation
L2 - Single RCT or Phase 2 Trials
Strong clinical evidence including:
- One completed Phase 3 RCT with positive results
- Multiple Phase 2 trials with consistent findings
Recommendation: Worth detailed feasibility assessment
L3 - Observational Studies
Supported by real-world evidence:
- Cohort studies
- Case-control studies
- Registry data
Recommendation: Consider with additional investigation
L4 - Preclinical/Mechanistic
Based on laboratory and mechanistic research:
- In vitro studies
- Animal models
- Pathway analysis
Recommendation: Explore as research hypothesis
L5 - Prediction Only
AI prediction without clinical validation:
- TxGNN model prediction
- May be KG, DL, or KG+DL validated
Recommendation: Use as starting point for research
Special Indicators
KG+DL (Dual Validated)
Predictions that appear in both:
- Knowledge Graph analysis
- Deep Learning model
These have higher confidence due to convergent evidence from independent methods.
High Score (>0.99)
Deep Learning predictions with confidence scores above 99% indicate very strong model certainty, though this doesn't guarantee clinical validity.
How to Use This Information
- Prioritize L1-L2 for immediate clinical relevance
- Investigate L3-L4 for emerging opportunities
- Use L5 with KG+DL as high-quality research hypotheses
- Always verify with current literature and clinical judgment
Remember: Evidence levels are guides, not guarantees. Clinical validation is always required.