文献类型：期刊文献

英文题名：Harnessing bioreactor heterogeneity: From gradient understanding to autonomous control via multiscale modeling and intelligent optimization

作者：Gu, Qingfeng[1];Yu, Junxiong[1];Liu, Yongqiang[1];Wang, Yongbo[1];Li, Chao[1,2];Zhuang, Yingping[1]

机构：[1]East China Univ Sci & Technol, State Key Lab Bioreactor Engn, 130 Meilong Rd, Shanghai 200237, Peoples R China;[2]Suzhou Womei Biol Co Ltd, Suzhou 215614, Peoples R China

年份：2026

卷号：90

外文期刊名：BIOTECHNOLOGY ADVANCES

收录：;EI(收录号：20261720578835);WOS:【SCI-EXPANDED(收录号:WOS:001756210400001)】；

基金：This work was supported by the National Natural Science Foundation of China [No. 22208099] , Shanghai Key Technology R & D Program "Synthetic Biology" Project [No. 25HC2820600] , and the Yangtze River Delta Community of Sci-Tech Innovation Joint R & D Projects [No. 2024CSJGG01800] .

语种：英文

外文关键词：Bioreactor scale-up; Multiscale modeling; Environmental gradients; Digital twin; Hybrid modeling; Autonomous biomanufacturing

摘要：The industrialization of biomanufacturing is constrained by the "scale-up effect", a phenomenon which is rooted in spatiotemporal heterogeneities that arise from multiscale interactions between hydrodynamics and cellular physiology in large-scale bioreactors. This review proposes a framework for a paradigm shift from passive observation to active, intelligent control. We first analyze how environmental gradients create distinct "cellular lifelines", which drive diverse physiological responses ranging from metabolic oscillations to population heterogeneity. We then demonstrate how multiscale modeling (which integrates computational fluid dynamics with physiological models) enables a strategic transition in scale-up strategy, shifting the focus from the futile elimination of gradients to their deliberate exploitation, with the ultimate aim of replicating a cell's critical environmental history. In addition, we explore the formation of a "mechanism-data symbiotic" hybrid modeling paradigm, wherein artificial intelligence enhances mechanistic foundations to facilitate real-time, adaptive optimization. Finally, we propose the digital twin as the ultimate embodiment of this evolution: a closed-loop autonomous system that transforms bioreactors from static vessels into cognitive entities capable of perception, learning, and self-optimization. While challenges in model generalizability and data integration remain, this roadmap points the way toward autonomous, efficient, and sustainable biomanufacturing.