AI时代中的电子显微学研究:严峻挑战、无穷机遇与壮阔前景  

Electron microscopy research in the age of AI:Severe challenges,Infinite opportunities,and Magnificent prospects

作　　者：刘铮 沈庆涛 隋森芳 LIU Zheng;SHEN Qingtao;SUI Senfang(Cryo-Electron Microscopy Center,Southern University of Science and Technology,Shenzhen Guangdong 518055;School of Life Sciences,Southern University of Science and Technology,Shenzhen Guangdong 518055;School of Life Science,Tsinghua University,Beijing 100084,China)

机构地区：[1]南方科技大学冷冻电镜中心,广东深圳518055 [2]南方科技大学生命科学学院,广东深圳518055 [3]清华大学生命科学学院,北京100084

出　　处：《电子显微学报》2025年第1期124-135,共12页Journal of Chinese Electron Microscopy Society

基　　金：国家自然科学基金专项资助项目(No.32241028).

摘　　要：近十年来结构生物学的发展突飞猛进,取得的标志性突破有两个,一是以冷冻电镜为代表的结构解析技术方向上的突破,二是以AlphaFold算法为代表的结构预测模型上的突破。2024年5月Google DeepMind更新了其生物结构预测工具,最新版本的AlphaFold 3拥有可以预测几乎所有分子类型的蛋白质复合体结构的能力,在药物互作预测方面也实现了很高的准确性。AlphaFold 3的发布为结构生物学研究带来巨大的变革,展现了AI技术的巨大潜力,也点燃了大众对生命科学和医学研究的热情与想象。与此同时,电子显微学的研究并未停下脚步,新技术、新方法层出不穷,在解析全新蛋白结构、超大超复杂复合体结构、动态结构、原位结构,以及更大尺度的细胞、组织、器官样品的研究中,电子显微学依旧有着不可替代的优势。当前有观点认为结构预测模型甚至可以替代以X射线晶体学和电子显微学为代表的传统实验科学,通过计算便能完成生物结构解析,但是这种观点是片面的。事实上,未来的结构生物学研究,必将是一个整合实验科学与AI技术,从单个蛋白或复合体的结构全面拓展到多蛋白复杂体系、细胞内原位、以及超越微观尺度进入到介观和宏观尺度等方面的研究。Over the past decade,structural biology has made significant achivements,there are two representational breakthroughs:First,in structural determination techniques,represented by Cryo-electron microscopy,and second,in structural prediction models,exemplified by the AlphaFold algorithm.In May 2024,Google DeepMind updated its biological structure prediction tool.The latest version,AlphaFold 3,possesses the ability to predict the structures of nearly all types of protein complexes and has also achieved high accuracy in predicting drug interactions.The release of AlphaFold 3 has brought about transformative changes structural biology,showcasing the immense potential of AI technology and fueling public enthusiasm and imagination for life science and biomedical research.Meanwhile,research in electron microscopy continues to advance with continuous emergence of new technologies and methods,electron microscopy science retains irreplaceable advantages in analyzing novel protein structures,extremely large and complicate complexes,dynamic structures,in⁃situ structures,and studies of larger scale samples including cells,tissues,and organs.Some propose that structural prediction models could even replace traditional experimental approaches such as X-ray crystallography and electron microscopy,allowing biomolecular structures analysis through computation alone.However,such opinions are biased,future research in structural biology will inevitably integrate experimental science and AI technology,expanding from the structures of individual proteins or complexes to comprehensive studies of multi⁃protein systems,intracellular in⁃situ locations,and beyond microscopic to mesoscopic and macroscopic scales.

关 键 词：电子显微学 高分辨结构 AlphaFold 结构模型预测 

分 类 号：Q6[生物学—生物物理学] Q71Q31Q51