基于分组贝叶斯排序的药物-靶标关系预测  

Drug-Target Interaction Prediction Based on Grouped Bayesian Ranking Approach

作　　者：丁棋梁 石泽智 李建华[1] DING Qiliang;SHI Zezhi;LI Jianhua(School of Information Science and Engineering,East China University of Science and Technology,Shanghai 200237,China)

机构地区：[1]华东理工大学信息科学与工程学院,上海200237

出　　处：《计算机工程与应用》2020年第15期185-190,共6页Computer Engineering and Applications

基　　金：国家科技重大专项项目(No.2018ZX09735002)。

摘　　要：基于贝叶斯排序的药物-靶标关系预测已经取得较好效果,但忽略了同一靶标的药物间的关联关系,从而影响精度。针对此问题,提出了一种新方法——基于分组贝叶斯排序的药物-靶标关系预测。在该方法中,根据与特定靶标存在相互作用的药物间具有相似性的现实,引入分组策略使这些相似药物间产生互动,并推导出基于分组策略的理论模型。该方法在五个公开数据集上与五种典型方法进行对比,产生的结果均优于所对比的方法。Drug-target interaction prediction based on Bayesian ranking has achieved good results,but neglects the correlation between drugs of the same target,which affects the accuracy.To solve this problem,a new method of drug-target interaction prediction based on grouped Bayesian ranking is proposed.In this method,according to the similarity of drugs interacting with specific targets,the grouping strategy is introduced to link these similar drugs,and the theoretical model based on grouping strategy is deduced.This method is comparing with five typical methods on five open datasets,and the experimental results are better than the comparison method.

关 键 词：药物-靶标相互作用 机器学习 贝叶斯排序 分组策略 

分 类 号：TP391.72[自动化与计算机技术—计算机应用技术]