TY - JOUR
T1 - A novel subnetwork representation learning method for uncovering disease-disease relationships
AU - Peng, Jiajie
AU - Guan, Jiaojiao
AU - Hui, Weiwei
AU - Shang, Xuequn
N1 - Publisher Copyright:
© 2020 Elsevier Inc.
PY - 2021/8
Y1 - 2021/8
N2 - Analyzing disease-disease relationships plays an important role for understanding disease mechanisms and finding alternative uses for a drug. A disease is usually the result of abnormal state of multiple molecular process. Since biological networks can model the interplay of multiple molecular processes, network-based methods have been proposed to uncover the disease-disease relationships recently. Given a disease and a network, the disease could be represented as a subnetwork constructed by the disease genes involved in the given network, named disease subnetwork. Because it is difficult to learn the feature representation of disease subnetworks, most existing methods are unsupervised ones without using labeled information. To fill this gap, we propose a novel method named SubNet2vec to learn the feature vectors of diseases from their corresponding subnetwork in the biological network. By utilizing the feature representation of disease subnetwork, we can analyze disease-disease relationships in a supervised fashion. The evaluation results show that the proposed framework outperforms some state-of-the-art approaches in a large margin on disease-disease/disease-drug association prediction. The source code and data are available athttps://github.com/MedicineBiology-AI/SubNet2vec.git.
AB - Analyzing disease-disease relationships plays an important role for understanding disease mechanisms and finding alternative uses for a drug. A disease is usually the result of abnormal state of multiple molecular process. Since biological networks can model the interplay of multiple molecular processes, network-based methods have been proposed to uncover the disease-disease relationships recently. Given a disease and a network, the disease could be represented as a subnetwork constructed by the disease genes involved in the given network, named disease subnetwork. Because it is difficult to learn the feature representation of disease subnetworks, most existing methods are unsupervised ones without using labeled information. To fill this gap, we propose a novel method named SubNet2vec to learn the feature vectors of diseases from their corresponding subnetwork in the biological network. By utilizing the feature representation of disease subnetwork, we can analyze disease-disease relationships in a supervised fashion. The evaluation results show that the proposed framework outperforms some state-of-the-art approaches in a large margin on disease-disease/disease-drug association prediction. The source code and data are available athttps://github.com/MedicineBiology-AI/SubNet2vec.git.
KW - Disease associations analysis
KW - Protein-protein interaction network
KW - Subnetwork representation learning
UR - http://www.scopus.com/inward/record.url?scp=85092611083&partnerID=8YFLogxK
U2 - 10.1016/j.ymeth.2020.09.002
DO - 10.1016/j.ymeth.2020.09.002
M3 - 文章
C2 - 32946974
AN - SCOPUS:85092611083
SN - 1046-2023
VL - 192
SP - 77
EP - 84
JO - Methods
JF - Methods
ER -