多特征提取和网络嵌入的蛋白质交互网络比对  

Aligning Protein-protein Interaction Networks Using Multi-feature Extraction and Network Embedding

作　　者：何浩发 马汇东 钟诚[1,2] HE Haofa;MA Huidong;ZHONG Cheng(School of Computer,Electronics and Information,Guangxi University,Nanning 530004,China;Key Laboratory of Parallel,Distributed and Intelligent Computing in Guangxi Universities and Colleges,Nanning 530004,China)

机构地区：[1]广西大学计算机与电子信息学院,南宁530004 [2]广西高校并行分布与智能计算重点实验室,南宁530004

出　　处：《小型微型计算机系统》2025年第4期847-855,共9页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(62362004,61962004)资助.

摘　　要：为快速获得高得分的拓扑和生物属性比对结果,提出一种利用生成式对抗网络的蛋白质交互网络全局比对高效方法.通过提取蛋白质网络中节点的序列比对得分、边的度、聚类系数和节点的Page Rank值,以组成节点的特征向量嵌入;将源网络的嵌入通过生成式对抗网络映射到目标网络的嵌入空间来得到初步比对结果;通过计算映射后嵌入向量之间的欧氏距离来获得最终比对结果.在真实的蛋白质交互网络数据集测试的结果表明,与已有的同类算法相比,在拓扑指标边正确性、诱导保守结构以及对称子图得分和基因同源一致性上,本文方法整体上具有优势.In order to quickly obtain alignment results with high-score topological and biological attribute,an efficient method for aligning global protein-protein interaction networks using generative adversarial networks is proposed.By extracting the sequence alignment score,edge degree,clustering coefficient,and Page Rank value of nodes in the network,the feature vector of the nodes is embedded.The embedding of the source network is mapped to the embedding space of the target network by generative adversarial network to obtain preliminary alignment results.The final alignment result is obtained by embedding the Euclidean distance between vectors after statistical mapping.The experimental results on the real protein-protein interaction network dataset show that compared with the existing methods,our proposed method overall has advantages for topological index edge correctness,induced conservative structure,symmetric subgraph score,and gene homology consistency.

关 键 词：蛋白质交互网络 网络比对 生成式对抗网络 网络嵌入 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]