BRAINCELL-AID: An Agentic AI Created Brain Cell Type Resource for Community Annotation
Researchers have developed BRAINCELL-AID, a novel multi-agent AI system designed to improve the annotation of brain cell types using single-cell RNA sequencing data. This system integrates free-text descriptions with ontology labels and employs retrieval-augmented generation (RAG) with PubMed literature to refine predictions and reduce hallucinations. BRAINCELL-AID achieved 77% accuracy in mouse gene set annotations and has been applied to over 5,000 brain cell clusters from the BRAIN Initiative Cell Census Network, yielding new insights into cell function and identifying Basal Ganglia-related cell types. AI
IMPACT Enhances biological research by improving cell type annotation accuracy and enabling new functional insights.