Research Cases

Example analyses demonstrating how to use SgTxGNN data for drug repurposing research.

---

Case Study 1: Metformin Beyond Diabetes

Background

Metformin is a first-line treatment for type 2 diabetes. SgTxGNN predictions suggest potential applications beyond glucose control.

SgTxGNN Predictions

Predicted Indication	Score	Source
Polycystic ovary syndrome	0.95	KG+DL
Colorectal cancer prevention	0.89	DL
Anti-aging	0.76	DL

Evidence Review

Polycystic Ovary Syndrome (PCOS)

• Clinical trials: Multiple RCTs support metformin for PCOS
• Mechanism: Improves insulin sensitivity, reduces androgens
• Status: Already used off-label in many countries

Cancer Prevention

• Observational studies show reduced cancer risk in diabetic patients on metformin
• Mechanism: AMPK activation, mTOR inhibition
• Status: Active clinical trials ongoing

Conclusion

This case demonstrates how TxGNN predictions align with emerging clinical evidence. Metformin's pleiotropic effects make it a strong repurposing candidate.

---

Case Study 2: Aspirin for Neurodegenerative Disease

Background

Aspirin is widely used for cardiovascular prevention. SgTxGNN suggests potential neurological applications.

SgTxGNN Predictions

Predicted Indication	Score	Source
Alzheimer's disease	0.82	KG
Parkinson's disease	0.71	DL

Evidence Review

Alzheimer's Disease

• Mechanism: Anti-inflammatory effects, reduction of amyloid aggregation
• Clinical evidence: Mixed results in RCTs
• Current status: Not recommended for prevention

Key Learning

High prediction scores don't guarantee clinical efficacy. Multiple failed trials highlight the importance of rigorous validation.

Conclusion

This case illustrates that even promising predictions require extensive validation. The gap between mechanistic plausibility and clinical benefit can be significant.

---

Case Study 3: Statins for Infection

Background

Statins are primarily used for cholesterol management. SgTxGNN identifies potential antimicrobial applications.

SgTxGNN Predictions

Predicted Indication	Score	Source
Sepsis	0.78	KG+DL
Pneumonia outcomes	0.72	DL

Evidence Review

Sepsis

• Mechanism: Immunomodulatory effects
• Observational data: Reduced mortality in statin users
• RCT results: SAILS trial showed no benefit

Research Implications

1. Observational benefits may not translate to RCT success
2. Patient selection may be critical
3. Timing of initiation matters

Conclusion

Demonstrates the complexity of translating predictions to clinical practice. Context and patient selection are crucial.

---

Tutorial: Analysing a Drug

Step 1: Find the Drug

import pandas as pd

# Load predictions
predictions = pd.read_csv('unified_predictions.csv')

# Filter for drug of interest
drug_preds = predictions[predictions['drug_name'] == 'Aspirin']
print(f"Found {len(drug_preds)} predictions for Aspirin")

Step 2: Review Predictions

# Sort by confidence
drug_preds = drug_preds.sort_values('score', ascending=False)

# Show top predictions
print(drug_preds[['disease_name', 'score', 'source']].head(10))

Step 3: Prioritise for Investigation

Focus on predictions that are:

• High confidence: Score > 0.9
• Dual validated: Source = KG+DL
• Clinically plausible: Has mechanistic rationale

Step 4: Gather Evidence

For each high-priority prediction:

1. Search ClinicalTrials.gov for existing trials
2. Review PubMed for mechanistic literature
3. Check for safety concerns

---

Best Practices

For Researchers

1. Start with high-confidence predictions
2. Verify mechanistic plausibility
3. Check for existing clinical evidence
4. Consider safety profile for new indication
5. Document your analysis process

For Clinicians

1. Predictions are hypotheses, not recommendations
2. Always verify with current literature
3. Consider patient-specific factors
4. Follow established clinical guidelines

---

Submit Your Case Study

Have an interesting analysis to share?

1. Document your methodology
2. Include reproducible code
3. Submit via GitHub Issues

Selected case studies may be featured on this page.

---

Educational Purpose
These case studies are for educational purposes only and do not constitute medical advice or treatment recommendations.