Flash-P

Evolving Agents for Knowledge Network Construction

An AI-powered framework for building, validating, and refining mechanistic causal biological networks through 5 specialized agents and deterministic validation.

Author & Institution

Christos Mitsanis
PhD Candidate
The University of Queensland
Brisbane, Australia
Computational Biology
AI Agents
Systems Biology

What is Flash-P?

Flash-P is a multi-agent framework that automates the construction and validation of mechanistic causal biological networks. Unlike static bioinformatics tools, Flash-P employs Evolving Agents — AI agents that evolve and improve through community input and expert curation.

5 Specialized Agents

  • CURATOR — Literature mining
  • PERTURBATION — Test case extraction
  • BUILDER — Network construction
  • VALIDATOR — Multi-method validation
  • REFINEMENT — Iterative improvement

3 Propagation Methods

  • Algebraic (Flash-P rules)
  • ODE with Hill functions
  • Random Walk with Restart

How to Cite

@article{mitsanis2026flashp,
  title={Flash-P: Evolving Agents for Automated Mechanistic Causal Network Construction},
  author={Mitsanis, Christos},
  journal={The University of Queensland},
  year={2026},
  note={Knowledge network traits Agent framework with 5 specialized AI agents}
}

License

Flash-P is released under the MIT License. You are free to use, modify, and distribute this software for academic and commercial purposes, with attribution.

Contributing

Flash-P thrives on community contributions. You can help by:

  • Building and validating networks with the desktop app
  • Reporting issues and suggesting improvements
  • Creating new validation plugins
  • Improving documentation

Technology Stack

Python 3.14
FastAPI
Next.js 15
React 19
TypeScript
Tailwind CSS
Cytoscape.js
Recharts
Framer Motion
shadcn/ui
Zustand
Claude Code