一种多视角图嵌入的miRNA-疾病关系预测方法  

Multi-view Graph Embedding Method for MiRNA-disease Association Prediction

作　　者：翁骞 孙宇平 凌捷[1] WENG Qian;SUN Yuping;LING Jie(School of Computer Science,Guangdong University of Technology,Guangzhou 510006,China)

机构地区：[1]广东工业大学计算机学院,广州510006

出　　处：《小型微型计算机系统》2025年第4期876-882,共7页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(62002070)资助.

摘　　要：MicroRNA(miRNA)在人类复杂疾病的发生发展中发挥着重要作用.由于生物学实验耗时费力,开发一种准确的计算预测方法对于识别疾病相关的miRNA是必不可少的.本文提出了一种多视角图嵌入的miRNA-疾病关联识别新方法(MGEMDA).该方法首先构建miRNA与疾病的多个相似网络,表征miRNA与疾病的关系.其次引入混合图表示学习框架,同时学习miRNA和疾病的特征.最后将特征输入到神经归纳矩阵补全模型中进行miRNA-疾病关联预测.在5折交叉验证中AUC值达到0.9454,与现有几种最先进的方法相比具有优异的性能.此外,对3种疾病(结肠癌、肺癌、乳腺癌)的案例研究进一步证实了MGEMDA对潜在疾病相关miRNA的预测能力.MicroRNA(miRNA)plays an important role in the development of human complex disease.As the biological experiments are time-consuming and labor-intensive,developing an accurate computational prediction method has become indispensable to identify disease-related miRNAs.We propose a novel method based on multi-view graph embedding for miRNA-disease associations identification(MGEMDA).Firstly,the multiple similarity networks for miRNAs and diseases are constructed to characterize the relationships of miRNAs and diseases.Secondly,the hybrid graph representation learning framework is introduced to learn the feature representations of miRNAs and diseases simultaneously.Finally,feature representations are input into a novel neural inductive matrix completion model to generate an association matrix.In the 5-fold cross validation,the AUC obtained by MGEMDA is 0.9454,demonstrating excellent performance compared to several existing state-of-the-art methods.Besides,case studies conducted on three diseases(colon cancer,lung cancer,breast cancer)further confirm the prediction capability of MGEMDA for predicting potential disease-related miRNAs.

关 键 词：miRNA-疾病关联预测 多视角相似性网络 图卷积网络 图注意力网络 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]