基于符号网络的药物靶标相互作用关系预测  

Drug-Target Interactions Prediction Based on Signed Networks

作　　者：雷涵清 蒋亚健 习智威 张万成 陈明 Lei Hanqing;Jiang Yajian;Xi Zhiwei;Zhang Wancheng;Chen Ming(School of Information Science and Engineering,Hunan Normal University,Changsha 410081)

机构地区：[1]湖南师范大学信息科学与工程学院,长沙410081

出　　处：《现代计算机》2022年第12期65-68,74,共5页Modern Computer

基　　金：湖南省教育厅科学研究项目(18C0016);湖南省大学生创新创业训练计划项目(S202010542075)。

摘　　要：药物-靶标相互作用(DTI)预测在新药物研发中具有重要意义。大多数计算方法将其建模为二元预测问题,忽视了DTI的具体类型。考虑DTI的积极或消极作用,将有利于研究多种药物对共同靶标的综合作用机理。通过构建药物靶标符号网络,将DTI预测问题转化为药物与靶标异构网络的符号链路预测问题,并引入Logistic回归与随机游走构建学习系统。在两个数据集进行实验,其预测结果呈现出良好的指标,表明该思路的可行性。Drug-target interaction(DTI)prediction plays an important role in the process of drug discovery.Most computa⁃tional methods model it as a binary prediction problem,ignoring the specific type of DTIs.Considering the positive or negative ef⁃fects of DTIs will facilitate the study of the comprehensive mechanism of multiple drugs on a common target.In this paper,we con⁃struct a drug-target signed network to transform the DTI prediction problem into a signed link prediction problem of drug-target heterogeneous network,and introduce the logistic regression and random walk to construct the learning system.Experiments are conducted on two datasets,and their prediction results show good performance in terms of indicators,indicating the feasibility of our idea.

关 键 词：药物-靶标相互作用 符号网络 链路预测 LOGISTIC回归 随机游走 

分 类 号：R96[医药卫生—药理学] O157.5[医药卫生—药学]