摘要
药物-靶标相互作用(DTI)鉴定是药物研发中的关键步骤,可有效缩小候选药物分子的搜索范围。同时,DTI鉴定也是多重药理和药物重定位等研究的基础。然而,通过生物实验研究DTI耗时长、成本高且伴有一定的盲目性。随着信息科学的飞速进步,人工智能(AI)在药物研发领域得到广泛应用,成为研究DTI的有效策略。根据算法设计原理的不同,用于DTI预测的AI方法可分为基于相似性、基于特征、基于网络和基于深度学习4类。本文重点介绍该4类方法的构建思路,并讨论模型评价问题和负样本问题。AI在DTI预测工作中具有巨大的发展潜力,可为药物研发带来新的机遇。
The identification of drug-target interactions(DTIs)is a critical step in drug discovery,which can effectively narrow down the candidate drug compounds to be searched for.Meanwhile,DTI prediction also underlies polypharmacology and drug repurposing.However,the study of DTIs through biological experiments is time-consuming,costly and not clearly-targeted.With the rapid development of information science,artificial intelligence(AI)has become a widely-used and effective strategy to study DTIs.Depending on the design principles of algorithms,AI methods for DTI prediction can be classified into four categories:similarity-based,feature-based,network-based and deep-learning methods.This paper introduces the construction of these methods and discusses the problems with model evaluation and negative samples.Overall,AI has a great potential in DTI prediction,which can bring new opportunities for drug research and development.
作者
李擎宇
张孝昌
王升启
LI Qing-yu;ZHANG Xiao-chang;WANG Sheng-qi(Institute of Radiation Medicine,Academy of Military Medical Sciences,Beijing 100850,China)
出处
《中国药理学与毒理学杂志》
CAS
北大核心
2022年第1期1-10,共10页
Chinese Journal of Pharmacology and Toxicology
基金
国家自然科学基金(81830101)。
关键词
药物-靶标相互作用
药靶组学
药物开发
人工智能
机器学习
drug-target interaction
targetomics
drug discovery
artificial intelligence
machine learning
