基于多标记学习预测药物-靶标相互作用  

Predicting drug-target interactions with multi-label learning

作　　者：彭利红[1] 刘海燕[1] 任日丽[1] 马俊[1] 王建芬[2] PENG Lihong;LIU Haiyan;REN Rili;MA Jun;WANG Jianfen(College of Information Engineering, Changsha Medical University, Changsha 410219, China;College of Pharmacy, Changsha Medical University, Changsha 410219, China)

机构地区：[1]长沙医学院信息工程学院,长沙410219 [2]长沙医学院药学院,长沙410219

出　　处：《计算机工程与应用》2017年第15期260-265,共6页Computer Engineering and Applications

基　　金：湖南省教育厅优秀青年项目(No.14B023);湖南省教育厅一般项目(No.13C1108;No.14C0115)

摘　　要：对药物-靶标关联进行了研究,提出基于弱标记和多信息融合的药物-靶标相互作用预测方法 PDML。通过与其他方法对比和数据库检索验证评估PDML模型的性能:与Yamanishi提出的方法、RLSMDA、Lap RLS及Net CBP相比,除在核受体数据集中该方法在AUC上的性能比Lap RLS略有降低之外,模型在敏感性、特异性、AUC和AUPR上的性能均优于其他四种方法;提取前5个预测分值最高的药物-靶标对,这些药物-靶标对能通过检索Drug Bank、Super Target和KEGG数据库而得到验证。Drug-target association prediction is researched.PDML based on weak label learning and multi-informationfusion is proposed to find new drug-target interactions from human enzymes,ion channels,GPCRs and nuclear receptors.The performance of the proposed method makes better than the methods provided by Yamanishi,RLSMDA,LapRLS andNetCBP in terms of sensitivity,specificity,AUC values and AUPR values except that the AUC values of the model slightlydecrease in nuclear receptor dataset compared to LapRLS.The five drug-target interaction pairs with highest scores can beextracted and validated by available public database DrugBank,SuperTarget and KEGG.

关 键 词：药物-靶标相互作用 多标记学习 多信息融合 药物-靶标相互作用网络 药物相似性 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]