《生命科学》 2025, 37(12): 1549-1562

细胞基础模型的领域进展

马英克^1,2 , 王思骐^3,4 , 李　鑫^3,4 , 冯桂海^3,4,*

¹国家生物信息中心，北京 100101 ²中国科学院北京基因组研究所，北京 100101 ³中国科学院动物研究所，北京 100101 ⁴北京干细胞与再生医学研究院，北京 100101

摘　要：

基于细胞基础模型表征能力对细胞状态进行预测和模拟，已成为当前生命科学领域的一种新型研究范式。该方法借助人工智能与大规模组学数据，旨在提取细胞相关特征并解码其状态调控网络，但在模型设计与构建方面其仍面临诸多挑战。本文系统梳理了细胞基础模型领域的关键进展，根据建模所使用数据类型，从单细胞组学、图像组学以及跨模态模型等方面，综述了各类具有代表性的模型及其建模思路。随后，本文概述了细胞基础模型在发展过程中于数据集构建、模型体系设计以及模型可解释性等方面面临的主要挑战。最后，从多个维度对未来发展方向进行了展望，期望推动跨物种、跨模态、跨尺度的细胞基础模型的构建，从而支撑生物基础研究与相关产业的创新发展。

通讯作者：冯桂海 , Email:fenggh@ioz.ac.cn

Advances in the field of cellular foundation models

MA Ying-Ke^1,2 , 李　鑫^3,4 , LI Xin^3,4 , FENG Gui-Hai^3,4,*

¹China National Center for Bioinformation, Beijing 100101, China ²Beijing Institute of Genomics, Chinese Academy of Sciences, Beijing 100101, China ³Institute of Zoology, Chinese Academy of Sciences, Beijing 100101, China ⁴Beijing Institute for Stem Cell and Regenerative Medicine, Beijing 100101, China

Abstract:

Prediction and simulation of cellular states based on the representational capacity of cellular foundation models has emerged as a new research paradigm in contemporary life sciences. This approach leverages artificial intelligence and large-scale omics datasets to extract cellular characteristics and decode regulatory network, but challenges remain in the design and implementation of foundation models. This review systematically summarizes key advances in the field of cellular foundation models. According to the data types used for model construction, we introduce representative models and modeling strategies based on single-cell omics, image-based omics, and cross-modal models. We then outline the major challenges in the development of cellular foundation models, including issues related to dataset construction, model architecture design, and model interpretability. Finally, we discuss future directions from multiple perspectives, with the goal of promoting the development of cellular foundation models that span species, modalities, and scales, thereby providing support for basic biological research and innovation in related industries.

Communication Author：FENG Gui-Hai , Email:fenggh@ioz.ac.cn