CellTypeAgent: Trustworthy cell type annotation with Large Language Models

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) enables high-resolution analysis of cellular heterogeneity.
• Accurate cell type annotation is essential for interpreting scRNA-seq data but remains challenging and time-consuming.
• Current annotation methods can suffer from inaccuracies and a tendency to hallucinate results.

Purpose of the Study:

• To introduce CellTypeAgent, a novel large language model (LLM)-agent designed for trustworthy cell type annotation in scRNA-seq data.
• To enhance the accuracy and reliability of cell type identification by integrating LLMs with database verification.
• To provide a more efficient and robust solution for a critical step in scRNA-seq data analysis.

Main Methods:

• Development of CellTypeAgent, an LLM-agent architecture.
• Integration of large language models with verification mechanisms querying relevant biological databases.
• Evaluation of CellTypeAgent performance across multiple real-world scRNA-seq datasets.

Main Results:

• CellTypeAgent demonstrated higher accuracy in cell type annotation compared to existing methods.
• The LLM-agent approach successfully mitigated the issue of hallucinations often associated with AI models.
• Consistent performance was observed across nine diverse datasets, covering 303 cell types from 36 human tissues.

Conclusions:

• CellTypeAgent offers a significant advancement in automated cell type annotation for scRNA-seq.
• The combination of LLMs and database verification provides a reliable and efficient tool for researchers.
• This approach holds substantial promise for accelerating discoveries in single-cell biology.