ATOM

Individual academic project · ULPGC · graded 9.4/10.

Re-engineering of a web platform that manages and runs automated omics analyses (RNA-seq), from a fragile monolith to a reproducible architecture with clear separation of concerns.

The previous version was a Flask monolith that stored data in JSON files and ran R Markdown scripts straight from the web server. It worked, but had serious limits in security, maintainability, scalability, traceability and reproducibility.

• Analyse and document the previous architecture and its risks.
• Separate the backend, the execution system and the analysis layer.
• Make executions reproducible and traceable.
• Improve security, data validation and user management.

• Analysis of the previous system and design of the new architecture.
• FastAPI backend: authentication, users, projects, entities, notifications and execution tracking.
• Next.js frontend with a dashboard, visual project management and reports.
• PostgreSQL data model and migration from JSON files.
• Asynchronous worker, Docker containers and continuous integration.

1. AMoving analysis execution out of the web server into an independent worker: the UI never blocks and every run is recorded with its state.
2. BPostgreSQL instead of JSON files for integrity, querying and a real history of projects and runs.
3. CDocker so the analysis environment, with its R and Python dependencies, is identical on any machine.
4. DAutomatic validation of experimental templates before an analysis starts, so it fails early and with a clear message.

A new architecture with a FastAPI backend, a Next.js frontend, PostgreSQL persistence, Docker orchestration and an independent worker process that runs analyses in the background.

Key features

• Omics project management
• Experimental template validation
• Asynchronous runs with live tracking
• HTML, DOCX and XLSX reports
• History and traceability
• Light and dark mode

• FastAPI
• Python
• Next.js
• TypeScript
• PostgreSQL
• Docker
• R Markdown
• GitHub Actions

• “Reproducibility is not an extra, but a basic condition of scientific software.” That was the guiding idea of the project.
• Rewriting well means first understanding why the old system worked, not only where it failed.

Private project: there is no public demo or open repository.