LLM4Cell: A Survey of Large Language and Agentic Models for Single-Cell Biology
Sajib Acharjee Dip, Adrika Zafor, Bikash Kumar Paul, Uddip Acharjee Shuvo, Muhit Islam Emon, Xuan Wang, Liqing Zhang
2025-10-13
Summary
This paper is a comprehensive overview of how new artificial intelligence tools, specifically large language models, are being used in the field of single-cell biology. It looks at different AI models designed to analyze complex biological data and how well they perform.
What's the problem?
Currently, the use of AI in single-cell biology is all over the place. Different researchers are using different types of data, building AI in different ways, and evaluating their results using different standards. This makes it hard to compare results and figure out which AI approaches are actually the best, and it hinders progress in the field.
What's the solution?
The researchers created a detailed survey called LLM4Cell, which examines 58 different AI models used for single-cell research. They grouped these models into categories based on how they work and what kind of data they analyze – things like RNA, DNA accessibility, and spatial information. They then tested these models on over 40 different datasets, looking at how well they perform on tasks like identifying cell types, predicting how cells will change, and forecasting drug responses. They also considered important factors like fairness, privacy, and how easily the models’ decisions can be understood.
Why it matters?
This work is important because it provides a unified view of the rapidly evolving field of AI in single-cell biology. By organizing and evaluating these different AI tools, it helps researchers understand what’s available, what works well, and where there are still challenges. It also highlights the need for better standards and more trustworthy AI development in this area, ultimately accelerating discoveries in biology and medicine.
Abstract
Large language models (LLMs) and emerging agentic frameworks are beginning to transform single-cell biology by enabling natural-language reasoning, generative annotation, and multimodal data integration. However, progress remains fragmented across data modalities, architectures, and evaluation standards. LLM4Cell presents the first unified survey of 58 foundation and agentic models developed for single-cell research, spanning RNA, ATAC, multi-omic, and spatial modalities. We categorize these methods into five families-foundation, text-bridge, spatial, multimodal, epigenomic, and agentic-and map them to eight key analytical tasks including annotation, trajectory and perturbation modeling, and drug-response prediction. Drawing on over 40 public datasets, we analyze benchmark suitability, data diversity, and ethical or scalability constraints, and evaluate models across 10 domain dimensions covering biological grounding, multi-omics alignment, fairness, privacy, and explainability. By linking datasets, models, and evaluation domains, LLM4Cell provides the first integrated view of language-driven single-cell intelligence and outlines open challenges in interpretability, standardization, and trustworthy model development.