《生命科学》 2025, 37(12): 1481-1492

生物医学大模型研究进展

张仪方¹ , 李　琛¹ , 程雨飞¹ , 张国庆² , 林新华^1,*

¹上海交通大学网络信息中心，上海 200240 ²中国科学院上海营养与健康研究所，上海 200032

摘　要：

从信息时代到智能时代，以大模型为代表的人工智能技术具有划时代意义，其发展迅猛，是全球科技研究的焦点。生物医学大模型(biomedical foundation models, BFMs) 作为人工智能与生物医学深度交叉的产物，依托于海量生物医学数据和大模型技术进步，通过整合多模态生物医学数据、创新算法架构与高性能算力，正在推动生命科学从传统实验驱动向“AI+”智能范式转型。本文梳理了生物医学大模型的核心数据构成、协同技术发展、多元化场景应用、带来的范式变革以及技术挑战难题和当前研究方向。生物医学大模型的持续进化有望实现从“辅助工具”到“智能协作者”的跃迁，引领生物医学传统研究工作的智能化变革。

通讯作者：林新华 , Email:james@sjtu.edu.cn

Advances in biomedical foundation models research

ZHANG Yi-Fang¹ , LI Chen¹ , CHENG Yu-Fei¹ , ZHANG Guo-Qing² , LIN Xin-Hua^1,*

¹Network & Information Center, Shanghai Jiao Tong University, Shanghai 200240, China ²Shanghai Institute of Nutrition and Health, Chinese Academy of Sciences, Shanghai 200032, China

Abstract:

From the information age to the intelligent age, artificial intelligence technology represented by large models has epoch-making significance, and its rapid development is the focus of global scientific and technological research. As a product of the deep intersection of artificial intelligence and biomedicine, biomedical foundation models, relying on massive biomedical data and large model technology progress, are promoting the transformation of life sciences from traditional experimental drive to the "AI+" intelligent paradigm by integrating multimodal biomedical data, innovative algorithm architecture and high-performance computing power. This article sorts out the core data composition, collaborative technology development, diversified scenario application, paradigm changes brought about by biomedical foundation models, technical challenges, and current research directions. The continuous evolution of biomedical foundation models is expected to leap from "auxiliary tools" to "intelligent collaborators", leading to the intelligent transformation of traditional biomedical research work.

Communication Author：LIN Xin-Hua , Email:james@sjtu.edu.cn