Large language model consensus substantially improves the cell type annotation accuracy for scRNA-seq data
- 1Department of Statistics, Texas A&M University, College Station, TX, USA
- 2Department of Statistics, Texas A&M University, College Station, TX, USA. [email protected]
- 3Division of Computational Biology, Department of Quantitative Health Sciences, Mayo Clinic, Rochester, MN, USA. [email protected]
Abstract
The rapid expansion of single-cell RNA sequencing (scRNA-seq) has made accurate cell type annotation a critical bottleneck for biological discovery. Existing computational methods are often limited by reference data dependency, while emerging single Large Language Model (LLM) approaches are susceptible to model-specific biases and provide insufficient uncertainty quantification. To address these limitations, we introduce mLLMCelltype, a framework that harnesses collective intelligence-the emergent problem-solving capacity arising when multiple independent agents interact through structured deliberation to produce solutions exceeding individual capabilities-of multiple LLMs through an iterative deliberation process. Across 49 diverse datasets, our framework achieves a mean accuracy of 77.2%, a 15.7-percentage-point improvement over the best-performing single-LLM baseline (61.5%). The consensus mechanism demonstrates high robustness to noisy input and generalizes to datasets released after the LLMs' training. By providing transparent reasoning chains and robust consensus-based confidence metrics, mLLMCelltype minimizes manual annotation effort and enables reliable interpretation of complex cellular landscapes. The framework is available as an open-source package and an accessible web server.
Related
- Combining CHARMM36m with OPC water improves accuracy
- Integration of exosomal SNORD46 and AL592064.2 with conventional tumor markers improves diagnostic accuracy for colorectal cancer
- ProteinJEPA: Latent prediction improves protein language model pretraining
- Efficient evidence-based genome annotation with EviAnn
- scBOA: Bayesian optimization of single-cell analysis pipelines for robust annotation and discovery
BibTeX
@article{chen2026large,
title = {{Large language model consensus substantially improves the cell type annotation accuracy for scRNA-seq data}},
author = {Yang Chen and Xianyang Zhang and Jun Chen},
journal = {Communications Biology},
year = {2026},
month = jun,
doi = {10.1038/s42003-026-10420-8},
url = {https://doi.org/10.1038/s42003-026-10420-8},
note = {CC BY-NC-ND},
}
RIS
TY - JOUR
TI - Large language model consensus substantially improves the cell type annotation accuracy for scRNA-seq data
AU - Yang Chen
AU - Xianyang Zhang
AU - Jun Chen
PY - 2026
DA - 2026/06/08
JO - Communications Biology
DO - 10.1038/s42003-026-10420-8
UR - https://doi.org/10.1038/s42003-026-10420-8
AB - The rapid expansion of single-cell RNA sequencing (scRNA-seq) has made accurate cell type annotation a critical bottleneck for biological discovery. Existing computational methods are often limited by reference data dependency, while emerging single Large Language Model (LLM) approaches are susceptible to model-specific biases and provide insufficient uncertainty quantification. To address these limitations, we introduce mLLMCelltype, a framework that harnesses collective intelligence-the emergent problem-solving capacity arising when multiple independent agents interact through structured deliberation to produce solutions exceeding individual capabilities-of multiple LLMs through an iterative deliberation process. Across 49 diverse datasets, our framework achieves a mean accuracy of 77.2%, a 15.7-percentage-point improvement over the best-performing single-LLM baseline (61.5%). The consensus mechanism demonstrates high robustness to noisy input and generalizes to datasets released after the LLMs' training. By providing transparent reasoning chains and robust consensus-based confidence metrics, mLLMCelltype minimizes manual annotation effort and enables reliable interpretation of complex cellular landscapes. The framework is available as an open-source package and an accessible web server.
SN - 2399-3642
ER -