bg-szy PDX Model Analysis Agent
The PDX Model Analysis Agent provides AI-driven analysis of patient-derived xenograft models for preclinical drug testing, translational research, and personalized oncology. It correlates PDX drug responses with patient outcomes and molecular profiles for treatment selection.
When to Use This Skill
- When selecting drug treatments based on PDX drug response data.
- To correlate PDX molecular profiles with patient tumor characteristics.
- For analyzing PDX-patient concordance in drug sensitivity.
- When designing preclinical drug combination studies.
- To identify biomarkers predicting PDX and patient drug response.
Core Capabilities
PDX-Patient Concordance: Analyze molecular similarity between PDX and donor tumor.
Drug Response Modeling: ML models correlating PDX drug sensitivity to patient outcomes.
Biomarker Discovery: Identify molecular features predicting drug response in PDX panels.
Combination Screening: Analyze synergy in PDX drug combination studies.
Translational Prediction: Project PDX findings to patient treatment selection.
Quality Assessment: Evaluate PDX fidelity and stability across passages.
PDX Quality Metrics
| Metric | Threshold | Interpretation |
|---|---|---|
| Genetic concordance | >90% | Variants maintained |
| Expression correlation | >0.85 | Transcriptome preserved |
| CNV fidelity | >85% | Copy number stable |
| Tumor take rate | Variable | Engraftment success |
| Passage stability | <P5 recommended | Minimal drift |
Workflow
Input: PDX molecular data, drug response curves, patient tumor data.
Concordance Analysis: Compare PDX to donor tumor at molecular level.
Drug Response Processing: Calculate IC50, AUC, TGI from growth curves.
Biomarker Analysis: Correlate molecular features with drug sensitivity.
Patient Prediction: Project findings to patient treatment recommendations.
Quality Assessment: Flag PDX models with significant drift.
Output: Drug rankings, biomarker associations, treatment recommendations.
Example Usage
User: "Analyze PDX drug response data for this breast cancer patient and recommend treatments."
Agent Action:
python3 Skills/Oncology/PDX_Model_Analysis_Agent/pdx_analyzer.py \
--pdx_rnaseq pdx_expression.tsv \
--pdx_mutations pdx_variants.maf \
--patient_tumor patient_expression.tsv \
--drug_responses pdx_drug_panel.csv \
--tumor_type breast_cancer \
--concordance_check true \
--output pdx_recommendations/
Drug Response Metrics
| Metric | Calculation | Interpretation |
|---|---|---|
| IC50 | Concentration for 50% inhibition | Potency |
| AUC | Area under dose-response curve | Overall sensitivity |
| TGI | Tumor growth inhibition % | In vivo efficacy |
| T/C | Treated/Control volume ratio | Treatment effect |
| Best response | Maximum tumor regression | Depth of response |
PDX Resource Integration
| Resource | Coverage | Data Types |
|---|---|---|
| PDXFINDER | 4000+ models | Multi-omic, drug response |
| PDMR (NCI) | 500+ models | Genomic, drug response |
| Champions/Crown | 1500+ models | Drug response |
| EurOPDX | 1000+ models | European cohort |
AI/ML Models
Drug Response Prediction:
- Gradient boosting on multi-omic features
- Gene expression signatures for drug classes
- Mutation-based response predictors
PDX-Patient Translation:
- Transfer learning from PDX to patient
- Domain adaptation for species differences
- Concordance-weighted predictions
Combination Synergy:
- Bliss independence model
- Loewe additivity analysis
- Machine learning synergy prediction
Clinical Translation Considerations
Factors Affecting Translation:
- Tumor heterogeneity: PDX from single biopsy
- Microenvironment: Mouse vs human stroma
- Immune system: Immunodeficient hosts
- Pharmacokinetics: Species differences
- Passage number: Drift over time
Best Practices:
- Use early passage PDX (P1-P5)
- Confirm molecular concordance
- Test drug at clinically-relevant doses
- Consider humanized PDX for immunotherapy
Prerequisites
- Python 3.10+
- scikit-learn, pandas
- Drug response databases
- PDX molecular datasets
Related Skills
- Drug_Repurposing - For alternative drug identification
- Multi_Omics_Integration - For PDX characterization
- Clinical_Trials - For trial matching
Output Report
- Concordance Summary: PDX-patient molecular similarity
- Drug Rankings: Predicted efficacy from PDX data
- Biomarker Associations: Features driving sensitivity
- Quality Flags: PDX reliability assessment
- Treatment Recommendations: Prioritized drug list
Author
AI Group - Biomedical AI Platform