MVGCN: data integration through multi-view graph convolutional network for predicting links in biomedical bipartite networks

被引:70
作者
Fu, Haitao [1 ]
Huang, Feng [1 ]
Liu, Xuan [1 ]
Qiu, Yang [1 ]
Zhang, Wen [1 ]
机构
[1] Huazhong Agr Univ, Coll Informat, Wuhan 430070, Peoples R China
基金
中国国家自然科学基金;
关键词
PROSTATE-CANCER CELLS; CLOZAPINE; PROLIFERATION;
D O I
10.1093/bioinformatics/btab651
中图分类号
Q5 [生物化学];
学科分类号
071010 ; 081704 ;
摘要
Motivation: There are various interaction/association bipartite networks in biomolecular systems. Identifying unobserved links in biomedical bipartite networks helps to understand the underlying molecular mechanisms of human complex diseases and thus benefits the diagnosis and treatment of diseases. Although a great number of computational methods have been proposed to predict links in biomedical bipartite networks, most of them heavily depend on features and structures involving the bioentities in one specific bipartite network, which limits the generalization capacity of applying the models to other bipartite networks. Meanwhile, bioentities usually have multiple features, and how to leverage them has also been challenging. Results: In this study, we propose a novel multi-view graph convolution network (MVGCN) framework for link prediction in biomedical bipartite networks. We first construct a multi-view heterogeneous network (MVHN) by combining the similarity networks with the biomedical bipartite network, and then perform a self-supervised learning strategy on the bipartite network to obtain node attributes as initial embeddings. Further, a neighborhood information aggregation (NIA) layer is designed for iteratively updating the embeddings of nodes by aggregating information from inter- and intra-domain neighbors in every view of the MVHN. Next, we combine embeddings of multiple NIA layers in each view, and integrate multiple views to obtain the final node embeddings, which are then fed into a discriminator to predict the existence of links. Extensive experiments show MVGCN performs better than or on par with baseline methods and has the generalization capacity on six benchmark datasets involving three typical tasks.
引用
收藏
页码:426 / 434
页数:9
相关论文
共 65 条
  • [1] Where to Position Clozapine: Re-examining the Evidence
    Agid, Ofer
    Foussias, George
    Singh, Shayna
    Remington, Gary
    [J]. CANADIAN JOURNAL OF PSYCHIATRY-REVUE CANADIENNE DE PSYCHIATRIE, 2010, 55 (10): : 677 - 684
  • [2] Ahmed A., 2013, WWW, P37
  • [3] Laplacian eigenmaps for dimensionality reduction and data representation
    Belkin, M
    Niyogi, P
    [J]. NEURAL COMPUTATION, 2003, 15 (06) : 1373 - 1396
  • [4] microRNA-155 promotes the proliferation of prostate cancer cells by targeting annexin 7
    Cai, Zhi-Kang
    Chen, Qi
    Chen, Yan-Bo
    Gu, Meng
    Zheng, Da-Chao
    Zhou, Juan
    Wang, Zhong
    [J]. MOLECULAR MEDICINE REPORTS, 2015, 11 (01) : 533 - 538
  • [5] Cao S, 2015, P 24 ACM INT C INF K, P891
  • [6] Chen X, 2019, IEEE INT C BIOINFORM, P354, DOI 10.1109/BIBM47256.2019.8983416
  • [7] LRSSLMDA: Laplacian Regularized Sparse Subspace Learning for MiRNA-Disease Association prediction
    Chen, Xing
    Huang, Li
    [J]. PLOS COMPUTATIONAL BIOLOGY, 2017, 13 (12)
  • [8] Datta SS, 2006, NEUROL INDIA, V54, P200
  • [9] Variational graph auto-encoders for miRNA-disease association prediction
    Ding, Yulian
    Tian, Li-Ping
    Lei, Xiujuan
    Liao, Bo
    Wu, Fang-Xiang
    [J]. METHODS, 2021, 192 : 25 - 34
  • [10] Network science of biological systems at different scales: A review
    Gosak, Marko
    Markovic, Rene
    Dolensek, Jurij
    Rupnik, Marjan Slak
    Marhl, Marko
    Stozer, Andra
    Perc, Matjaz
    [J]. PHYSICS OF LIFE REVIEWS, 2018, 24 : 118 - 135