固定骨架的从头蛋白质设计:多目标优化与深度学习算法研究进展
Advances in computational de novo protein design with fixed backbone based on muti-objective optimization and deep learning
李瑞祥 1沈红斌1
作者信息
- 1. 上海交通大学图像处理与模式识别研究所,上海 200240
- 折叠
摘要
固定骨架的蛋白质设计通过计算的方法生成能够折叠为目标蛋白结构的氨基酸序列,该过程可以被视作蛋白质结构预测的逆过程.蛋白质的功能与结构有着紧密的联系,因此基于特定结构的蛋白质设计在酶、疫苗、制药、蛋白质材料等领域都有着潜在的作用.基于蛋白质设计的方法原则,根据当前该领域的进展讨论了基于能量函数优化和基于深度学习这两种主要的蛋白质设计算法类型,最后总结当前蛋白质设计领域的瓶颈问题并做出了展望.
Abstract
Fixed backbone protein design generates amino acid sequences capable of folding into target protein structures by computational methods,which can be regarded as the inverse process of protein structure predic-tion.The function of proteins is closely linked to their structure;hence,protein design based on specific struc-tures plays a potential pivotal role in fields such as enzymology,vaccines,pharmaceuticals,and protein materi-als.This paper briefly introduces the principles of protein design methods and then,based on current progress in the field,discusses two main types of protein design algorithms:those based on energy function optimization and those based on deep learning.Finally,we summarize the bottleneck in the field of protein design and discuss the potential directions in this field.
关键词
蛋白质设计/能量函数/多目标优化/深度学习/蛋白质序列与结构Key words
protein design/energy function/multi-objective optimization/deep learning/protein sequence and structure引用本文复制引用
出版年
2024
NETL
NSTL
万方数据