RDRGSE: A Framework for Noncoding RNA-Drug Resistance Discovery by Incorporating Graph Skeleton Extraction and Attentional Feature Fusion

被引：1

作者：

Zhang, Ping ^{[1
]}

Wang, Zilin ^{[1
]}

Sun, Weicheng ^{[1
]}

Xu, Jinsheng ^{[1
]}

Zhang, Weihan ^{[1
]}

Wu, Kun ^{[4
]}

Wong, Leon ^{[2
,3
]}

Li, Li ^{[1
,5
]}

机构：

[1] Huazhong Agr Univ, Coll Informat, Hubei Key Lab Agr Bioinformat, Wuhan 430070, Peoples R China

[2] Guangxi Acad Sci, Guangxi Key Lab Human Machine Interact & Intellige, Nanning 530007, Peoples R China

[3] Tongji Univ, Inst Machine Learning & Syst Biol, Sch Elect & Informat Engn, Shanghai 200092, Peoples R China

[4] Univ Calif Riverside, Dept Biochem, Riverside, CA 92521 USA

[5] Huazhong Agr Univ, Hubei Hongshan Lab, Wuhan 430070, Peoples R China

来源：

ACS OMEGA | 2023年 / 8卷 / 30期

基金：

中国国家自然科学基金;

关键词：

RESOURCE;

D O I：

10.1021/acsomega.3c02763

中图分类号：

O6 [化学];

学科分类号：

0703 ;

摘要：

Identifying noncoding RNAs (ncRNAs)-drug resistance associationcomputationally would have a marked effect on understanding ncRNAmolecular function and drug target mechanisms and alleviating thescreening cost of corresponding biological wet experiments. Althoughgraph neural network-based methods have been developed and facilitatedthe detection of ncRNAs related to drug resistance, it remains a challengeto explore a highly trusty ncRNA-drug resistance association predictionframework, due to inevitable noise edges originating from the batcheffect and experimental errors. Herein, we proposed a framework, referredto as RDRGSE (RDR association prediction by using graph skeleton extractionand attentional feature fusion), for detecting ncRNA-drug resistanceassociation. Specifically, starting with the construction of the originalncRNA-drug resistance association as a bipartite graph, RDRGSE tookadvantage of a bi-view skeleton extraction strategy to obtain twotypes of skeleton views, followed by a graph neural network-basedestimator for iteratively optimizing skeleton views aimed at learninghigh-quality ncRNA-drug resistance edge embedding and optimal graphskeleton structure, jointly. Then, RDRGSE adopted adaptive attentionalfeature fusion to obtain final edge embedding and identified potentialRDRAs under an end-to-end pattern. Comprehensive experiments wereconducted, and experimental results indicated the significant advantageof a skeleton structure for ncRNA-drug resistance association discovery.Compared with state-of-the-art approaches, RDRGSE improved the predictionperformance by 6.7% in terms of AUC and 6.1% in terms of AUPR. Also,ablation-like analysis and independent case studies corroborated RDRGSEgeneralization ability and robustness. Overall, RDRGSE provides apowerful computational method for ncRNA-drug resistance associationprediction, which can also serve as a screening tool for drug resistancebiomarkers.

引用

页码：27386 / 27397

页数：12