基于网络表示学习的miRNA功能相似性研究  

Research on Functional Similarity of miRNA Based on Network Representation Learning

作　　者：阮璐 熊赟[1] RUAN Lu;XIONG Yun(School of Computer Science,Fudan University,Shanghai 201203,China)

机构地区：[1]复旦大学计算机科学技术学院,上海201203

出　　处：《计算机工程》2019年第2期154-159,共6页Computer Engineering

基　　金：国家高技术研究发展计划(2015AA020105-10);上海市科委基金(16JC1400800;17511105502)

摘　　要：miRNA是一类重要的非编码小RNA分子,与癌症等疾病有密切的关系。目前研究者已经识别大量miRNA,但是多数miRNA的功能仍然未知。为此,提出一种网络表示学习的miRNA功能相似性计算方法。通过miRNA的相关数据集如目标基因和关联疾病可以有效地计算miRNA的功能相似性,从而预测疾病相关的候选miRNA。利用不同类型生物数据集构建miRNA相关多源网络,采用网络表示学习的方式为网络中的每一个miRNA节点学习一个特征向量,进而使用特征向量来衡量miRNA的相似性。实验结果表明,与DeepWalk方法相比,该方法在同一家族的miRNA中能够取得较高的得分,并且可以在已有的数据库中找到疾病候选miRNA验证记录。miRNA is an important class of non-coding small RNA molecules that are closely related to diseases such as cancer.Researchers have identified a large number of miRNA,but the function of most miRNA remains unknown.Therefore,a method for calculating the similarity of miRNA functions based on network representation learning is proposed.The functional similarity of miRNA can be efficiently calculated by relevant data sets of miRNA such as target genes and associated diseases,thereby predicting disease-related candidate miRNA.miRNA-related multi-source network is constructed by utilizing different types of biological data sets.The network representation learning method is used to learn a feature vector for each miRNA node in the network,and the similarity of the miRNA is measured by the learned feature vector.Experimental results show that compared with the DeepWalk method,the method can obtain higher scores in the same family of miRNA,and the disease candidate miRNA verification records can be found in the existing database.

关 键 词：功能相似性 网络表示学习 网络嵌入 多源网络 特征向量 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]