Technologies

Rust Engine (gpredomics)

The core ML engine is implemented in Rust for maximum performance:

  • Rust – All evolutionary algorithms (GA, Beam, MCMC), fitness evaluation, cross-validation, and jury voting
  • CUDA – Optional GPU acceleration for fitness evaluation on large datasets
  • PyO3 / Maturin – Python bindings exposing the full API as a native Python module (gpredomicspy)
  • Serde / YAML – Parameter serialization and configuration loading
  • Rayon – Parallel computation for population evaluation

Performance: up to 1,000x speedup over the original R implementation, processing thousands of features across hundreds of samples in seconds.

Web Application (predomicsapp-web)

Backend (Python)

  • FastAPI – Async REST API with automatic OpenAPI documentation (50+ endpoints)
  • SQLAlchemy 2.0 – Async ORM with PostgreSQL for user, project, dataset, and job management
  • Pydantic v2 – Request/response validation and serialization
  • WebSocket – Real-time job monitoring with structured progress streaming
  • reportlab – PDF biomarker report generation
  • nbformat – Jupyter notebook (.ipynb) generation
  • JWT + API Keys – Authentication with role-based access control
  • Background Tasks – Non-blocking job execution with subprocess management

Frontend (JavaScript)

  • Vue.js 3 – Reactive UI with Composition API and <script setup> syntax
  • Pinia – State management for auth, projects, and application state
  • Vue Router – Client-side routing with authentication guards
  • vue-i18n – Full internationalization (English and French)
  • Plotly.js – All interactive charts (AUC evolution, heatmaps, violin plots, SHAP, waterfall, etc.)
  • Vite – Fast development server and optimized production builds
  • IntersectionObserver – Scroll-reveal animations on the landing page

Infrastructure

  • Docker Compose – Single-command deployment with multi-stage builds
  • PostgreSQL 16 – Production database with migration support
  • NGINX – Reverse proxy for static assets and API routing
  • pytest + Vitest – Backend and frontend test suites

R Package (predomicspkg)

The original implementation and reference for the algorithms:

  • R – Core algorithm implementations (TerGa1, TerGa2, TerBeam, TerDa)
  • doSNOW / foreach – Parallel computation across multiple cores
  • pROC – ROC curve analysis and AUC computation
  • glmnet – Regularized regression for comparison baselines
  • viridis – Publication-quality color palettes for visualizations
  • Shiny – Original interactive web interface (legacy, superseded by PredomicsApp)

GitHub Pages (predomics.github.io)

  • Jekyll – Static site generator with the just-the-docs theme
  • Markdown – Documentation content
  • Vanilla JS – Scroll animations, screenshot gallery, counter effects
This site uses Just the Docs, a documentation theme for Jekyll.