PhenoGenX (PGX)
Dual-Engine Platform for HIV-1 Drug Resistance
PhenoGenX (PGX) is a dual-engine, data-driven HIV-1 drug resistance interpretation system integrating curated rule-based algorithms with machine learning–based phenotypic prediction models. The platform is designed to enhance molecular resistance interpretation, improve fold-change estimation, and support research, surveillance, and treatment optimization.
Platform Overview
• Established rule-based mutation interpretation frameworks
• Ensemble machine learning models trained on large-scale genotype–phenotype datasets
• Quantitative fold-change prediction across major antiretroviral classes
The system is designed for research-grade analysis and advanced clinical resistance interpretation.
Analytical Architecture
Rule-Based Engine
• Syndemic vulnerability index (continuous & categorical)
• Country-level risk gradients
• Subnational risk mapping
• Demographic stratification (age, sex, wealth, region)
• Risk-adjusted HIV prevalence modeling
• Temporal evolution analysis
Machine Learning Engine
- Trained on >40,000 genotype–phenotype pairs
- Predicts phenotypic fold-change (IC₅₀)
- Cross-validated ensemble modeling
- Optimized using CRPS, R², MAE, and calibration metrics
Data Foundation
Rule-Based Engine
Each isolate includes:
Amino acid mutation profile of PR, RT, and IN regions relative to HXB2
Associated phenotypic susceptibility expressed as fold-change
Multi-class drug coverage (NRTI, NNRTI, PI, INSTI)
Validation and Performance
Rule-Based Engine
PGX performance has been evaluated using:
• Cross-validation
• External dataset benchmarking
• Concordance comparison with established interpretation systems
• Calibration analysis of fold-change prediction
Detailed validation results are described in the PGX manuscript.
Applications
Rule-Based Engine
- Third-line ART failure investigation
- Surveillance of transmitted drug resistance
- Minority variant analysis
- Quantitative resistance modeling
- Research and digital epidemiology applications
