About
GeneAgent is a first-of-its-kind language agent built upon GPT-4 to automatically interact with domain-specific databases for gene set annotation. It generates interpretable and contextually accurate biological process names for user-provided gene sets, either aligning with significant enrichment analyses or introducing novel terms. At its core is a self-verification mechanism that autonomously interacts with various expert-curated biological databases through Web APIs, performing fact verification and providing objective evidence to support or refute raw LLM output, reducing hallucination and enabling reliable evidence-based insights.
Tech Stack
PythonGPT-4Azure OpenAIPyTorchPandasRequests
Research Paper
DOI: 10.5281/zenodo.15008591Quick Start
git clone git@github.com:ncbi-nlp/GeneAgent.git && conda create -n geneagent python=3.11 && pip install openai torch numpy pandas requests && python main_cascade.py