PhenoGenX Documentation

Comprehensive technical documentation, implementation guides, and scientific specifications for the PhenoGenX HIV drug resistance interpretation platform.

Technical Specifications
System Architecture
  • Backend: Flask + Python 3.10+
  • Frontend: Bootstrap 5 + JavaScript
  • ML Framework: Scikit-learn, LightGBM, XGBoost
  • Database: SQLite/PostgreSQL
API Documentation
  • RESTful API endpoints
  • JSON request/response formats
  • Authentication & rate limiting
View Technical Docs
Database & Schemas
Database Schema
  • Sequence metadata storage
  • Mutation profile tables
  • Prediction results schema
  • User & session management
Data Models
  • ER Diagrams
  • Primary/Foreign key relationships
  • Data migration scripts
View Schema Docs
SOPs & Protocols
Standard Operating Procedures
  • Sequence quality control SOP
  • ML model validation protocol
  • Data security & privacy SOP
  • Report generation workflow
Implementation Guides
  • Deployment checklist
  • Maintenance procedures
  • Troubleshooting guide
View SOPs
Technical Implementation Details

HXB2 Alignment Pipeline

Multi-stage alignment process for accurate mutation calling:

  1. MAFFT Alignment: Initial sequence alignment using MAFFT v7.525
  2. Muscle Refinement: MUSCLE v5.1 for consensus refinement
  3. Biopython Fallback: Custom Python alignment for edge cases
  4. HXB2 Coordinate Mapping: Standardized position mapping
Mutation Calling Logic
  • Amino acid substitution detection
  • Insertion/deletion handling
  • Ambiguous base resolution
  • Quality score integration

Algorithm Architecture

Multi-source rule integration for comprehensive resistance interpretation:

  • Stanford HIVDB: Weighted mutation scoring system
  • ANRS Algorithm: French National Agency scoring
  • IAS-USA Guidelines: Major/accessory mutation classification
  • WHO SDRM: Surveillance drug resistance mutations
Scoring System
Resistance Level Score Range Clinical Interpretation
Susceptible 0-9 No significant resistance
Potential Low-Level 10-14 Possible reduced susceptibility
Low-Level 15-29 Reduced susceptibility
Intermediate 30-59 Significant resistance
High-Level ≥60 High-level resistance

CRPS-Optimized Ensemble

Continuous Ranked Probability Score optimization for model selection:

Base Models
  • ElasticNet: Regularized linear regression
  • LightGBM: Gradient boosting framework
  • XGBoost: Extreme gradient boosting
  • Random Forest: Ensemble decision trees
Optimization
  • CRPS-based weighting
  • Cross-validation (5-fold)
  • Hyperparameter tuning
  • Model calibration
Training Dataset
  • Size: 45,000+ genotype-phenotype pairs
  • Sources: Stanford, Los Alamos, EPHI clinical data
  • Subtypes: A, B, C, CRF01_AE, CRF02_AG
  • Drugs: 22 ARV medications

Supported Input Formats
FASTA Format
>Sequence_ID
ATGACC...
Mutation CSV
ID,Mutations
SEQ001,K103N,M184V
Plain Mutation List
K103N, M184V, G190A
Output Formats
  • JSON: Complete structured response
  • CSV (Wide): One row per sequence
  • CSV (Long): One row per sequence-drug pair
  • PDF Reports: Clinical summary reports
  • Excel: Formatted worksheets
Release Notes & Changelog
Version History
v1.0 (Current)

Initial production release with core rule-based and ML engines.

v1.1 (Planned)

Batch processing enhancements and API improvements.

v1.2 (Planned)

Subtype inference and advanced visualization tools.

v2.0 (Roadmap)

Full API ecosystem and integration capabilities.

Full Changelog
Summary Reports & Analytics
Platform Analytics
  • Usage statistics dashboard
  • Data processing metrics
  • Performance benchmarks
  • User activity reports
Validation Reports
  • Model performance validation
  • Concordance analysis reports
  • Monthly summary reports
View Reports
Documentation Status

Complete: Technical specifications, API documentation, database schemas

In Progress: SOPs, implementation guides, detailed protocol documentation

Planned: Interactive API playground, video tutorials, comprehensive PDF manual (v1.3)