tnbc-drug-repurposing

TNBC Drug Repurposing: Brigatinib as a Novel Candidate

Executive Summary

This project identifies Brigatinib (Alunbrig), an FDA-approved ALK/EGFR inhibitor for non-small cell lung cancer, as a promising drug repurposing candidate for triple-negative breast cancer (TNBC). Using machine learning-based drug-target interaction prediction validated against the DepMap PRISM drug screening dataset, we demonstrate that Brigatinib exhibits superior TNBC sensitivity compared to all other FDA-approved EGFR inhibitors.

---

Key Findings

1. Brigatinib Ranks #1 Among EGFR Inhibitors for TNBC Sensitivity

2. Statistical Validation

• Head-to-head comparison: Brigatinib outperforms Erlotinib in 10/11 TNBC cell lines (paired t-test p=0.0006)
• TNBC selectivity: +58.4% differential sensitivity (TNBC vs non-TNBC breast cancer)
• Correlation with predictions: Pearson r=0.495 (p=0.01) between ML predictions and DepMap validation

3. Sensitive TNBC Cell Lines

---

Methods

Data Sources

1. ChEMBL Database: Drug-target interaction data for model training
2. DepMap PRISM: Primary drug screening dataset (4,688 compounds, 578 cell lines)
3. Breast Cancer Cell Lines: 26 lines (12 TNBC, 7 ER+, 4 HER2+, 2 Triple-positive)

Machine Learning Pipeline

01_fetch_chembl.py      → Download ChEMBL bioactivity data
02_clean_data.py        → Process and filter IC50 values
03_featurize.py         → Generate molecular descriptors (Morgan fingerprints)
04_train_models.py      → Train XGBoost classifiers for EGFR, PARP1, PIK3CA
05_rank_compounds.py    → Score drug library compounds
06_screen_compounds.py  → Prioritize high-confidence hits
07_screen_drug_library.py → Screen FDA-approved drugs
08_depmap_validation.py → External validation against DepMap

Validation Strategy

1. Internal validation: 5-fold cross-validation (ROC-AUC > 0.85)
2. External validation: DepMap PRISM drug sensitivity data
3. Competitive analysis: Head-to-head comparison with approved EGFR inhibitors

---

Project Structure

brca/
├── data/
│   ├── drug_repurposing/       # Drug library and screening results
│   ├── processed/              # Featurized data and model inputs
│   ├── screening_results/      # Benchmark and validation outputs
│   └── validation/             # DepMap validation results
├── depmap/                     # DepMap PRISM raw data
├── models/                     # Trained model files and cards
├── notebooks/
│   ├── depmap_validation.ipynb
│   └── brigatinib_analysis.ipynb
└── scripts/                    # Pipeline scripts (01-08)

---

Reproducibility

Environment Setup

python -m venv .venv
source .venv/bin/activate
pip install pandas numpy scikit-learn xgboost rdkit matplotlib seaborn scipy

Running the Pipeline

# Full pipeline
python scripts/01_fetch_chembl.py
python scripts/02_clean_data.py
python scripts/03_featurize.py
python scripts/04_train_models.py
python scripts/05_rank_compounds.py
python scripts/06_screen_compounds.py
python scripts/07_screen_drug_library.py
python scripts/08_depmap_validation.py

Running Validation Analysis

jupyter notebook notebooks/brigatinib_analysis.ipynb

---

Clinical Relevance

Why Brigatinib for TNBC?

1. Unmet medical need: TNBC lacks targeted therapies; only chemotherapy is standard of care
2. Multi-kinase activity: Brigatinib inhibits ALK, EGFR, IGF-1R, and FLT3
3. FDA-approved: Already approved for ALK+ NSCLC, enabling faster clinical translation
4. Novel indication: Only 10 PubMed results for "brigatinib breast cancer" (as of Dec 2024)

Proposed Mechanism

Brigatinib's superior TNBC activity may be attributed to:

• Dual ALK/EGFR inhibition with broader kinase coverage
• Activity against IGF-1R, which is overexpressed in TNBC
• Different binding mode compared to first-generation EGFR inhibitors

---

Limitations

1. In vitro data only: DepMap uses cell line viability, not tumor models
2. Sample size: 11-12 TNBC cell lines available for analysis
3. Single dose: PRISM uses standardized dosing, not dose-response curves
4. Biomarker unknown: Molecular predictors of response not yet identified

---

Future Directions

1. Biomarker discovery: Correlate sensitivity with genomic features (BRCA status, EGFR expression)
2. Combination studies: Explore Brigatinib + PARP inhibitor synergy in BRCA-mutant TNBC
3. Preclinical validation: Patient-derived xenograft (PDX) studies
4. Clinical trial design: Phase II basket trial in TNBC

---

References

1. DepMap PRISM Repurposing Dataset: https://depmap.org/portal/prism/
2. Brigatinib FDA Label: https://www.accessdata.fda.gov/drugsatfda_docs/label/2020/208772s008lbl.pdf
3. ChEMBL Database: https://www.ebi.ac.uk/chembl/

---

Authors

TNBC Drug Repurposing Project

License

MIT License

---

Acknowledgments

• Broad Institute DepMap for PRISM screening data
• ChEMBL for bioactivity data
• RDKit for molecular featurization