生物加工过程2025，Vol.23Issue(6)：628-638,11.DOI:10.3969/j.issn.1672-3678.2025.06.004
小样本机器学习在酶工程中的应用进展
Advances in the application of few-shot learning in enzyme engineering
摘要
Abstract
Machine learning,as an emerging powerful tool in enzyme engineering,can enable the elucidation of complex relationships between biological sequences and functions,thereby accelerating the identification and design of high-performance enzymes.However,this approach heavily relies on large volume of high-quality labeled data,posing significant challenges with regards to data acquisition via wet-lab experiments.Recently,few-shot learning,particularly through transfer learning strategies,provided novel solutions to the data scarcity issue and demonstrated great potential in enzyme engineering.In this review,we first outlined the typical workflow of applying machine learning to enzyme engineering—from dataset construction and feature extraction to model training,functional prediction,and experimental validation.Then,we highlighted recent advances in using few-shot learning for optimizing enzyme activity,substrate specificity,and stereoselectivity.In the end,we presented future research directions in this field.关键词
小样本学习/机器学习/酶工程/定向进化Key words
few-shot learning/machine learning/enzyme engineering/directed evolution分类
生物科学引用本文复制引用
周佳楠,杨立荣,江玲,于浩然..小样本机器学习在酶工程中的应用进展[J].生物加工过程,2025,23(6):628-638,11.基金项目
浙江省"尖兵""领雁"科技计划(2025C01097) （2025C01097）
