Skip to content

About

No description, website, or topics provided.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Latest commit

 

History

174 Commits

Folders and files

Repository files navigation

RNADesignAgent

A tool-first, calibrated multi-agent system for RNA bioinformatics design and analysis. The language model orchestrates, reviews and reports — every designed molecule traces to a deterministic generator call, and every prediction claim is checked against the measured capability of the tool that produced it.

The system packages 41 scientific tools (20 vendored external repositories) behind local microservices and exposes them through 11 domain workflows: mRNA design (CDS and full-length), siRNA and ASO design, sgRNA design, RNA structure prediction and inverse folding (secondary and tertiary), aptamer design, RNA–ligand and RNA–protein interaction scoring, and IRES prediction.

Quick start

git clone --recurse-submodules https://github.com/xlab-BioAI/RNADesignAgent.git
cd RNADesignAgent/rna_agent_system
pip install -r requirements.txt
PYTHONPATH="$(pwd):$PYTHONPATH" python cli/form_chat.py

The CLI walks you through a resource check, workflow selection, automatic microservice startup and a structured per-workflow form. Three submodules are pinned to patched forks and several services need manually placed model weights — see the deployment guide before your first run.

Documentation

All documentation lives under rna_agent_system/docs/:

Document Purpose
docs/README.md Project overview and quick start
docs/DEPLOYMENT.md Full environment setup, weights, compilation, curl examples
docs/WORKFLOWS.md The 11 domain workflows in depth
docs/TOOLS.md All 41 tools, port registry, microservice notes
docs/ARCHITECTURE.md Seven-layer architecture and core module APIs

Benchmarks and data

The benchmark datasets used in the paper are included under rna_agent_system/data/benchmarks/ (210-task platform benchmark, 31-session process-aware benchmark, and the mRNA/aptamer target sets). The frozen final-form datasets, the external anchor sets, the packaged capability map and the model-usage register are deposited at Zenodo under DOI 10.5281/zenodo.22736194. See docs/README.md §8 for how to reproduce a benchmark task through the CLI.

License

Released under the license in LICENSE.

About

No description, website, or topics provided.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages