基于多重相似性和增强注意力预测药物-靶标相互作用

浏览次数:10
  • 分享到:

摘要:

在新药发现和药物重定位研究中,发现药物与靶标之间的相互作用是重要的研究内容。针对药物与靶标相互作用网络,提出一种基于多重相似性和增强注意力机制的图卷积神经网络模型 (RSGCN) 预测药物靶标相互作用。首先,提出了多重相似性来捕捉网络结构特征,以充分利用节点间的直接或间接关系。然后,通过 PCA 降维减少相似性噪声对实验结果的影响。最后,采用图卷积神经网络(graph convolution neural network,GCN) 获得节点嵌入表示,并融入基于注意力的增强层,通过增强注意力机制获得节点间的注意力权重,能够高效地预测药物与靶标之间的相互作用。在黄金标准数据集上的实验结果表明 RSGCN 模型具有较好的性能。

In the research of new drug discovery and drug repositioning, it is important to search for the interactions between drugs and targets. In this study, we propose a graph convolutional network model based on multiple similarities and enhanced attention mechanism to predict drugtarget interactions(RSGCN) for the drug-target interaction network, Firstly, we propose to use multiple similarities to optimize the original feature vectors of drugs and targets, capture network structure features, and fully utilize direct or indirect relationships betwen nodes, Then, we reduce the impact of similarity noise on experimental results through PCA dimensionality reduction, Finally, we use GCN to obtain node embedding representations and incorporate an attention-based enhanced layer to obtain attention weights between nodes. which efficiently predicts interactions between drugs and targets, The experiments use a public gold standard dataset, and the experimental results indicate that the RSGCN model has good performance.

作者:

王伟,余梦雪,孙斌,万仕彤,刘栋,周运,张红军,王鲜芳

Wang Wei,Yu Mengxue,Sun Bin,Wan Shitong, Liu Dong,Zhou Yun,Zhang Hongjung,Wang Xianfang

机构地区:

计算机与信息工程学院;河南省教育人工智能与个性化学习重点实验室;河南理工大学鹤壁工程技术学院;河南工学院 计算机科学与技术学院

引用本文:

王伟、余梦雪、孙斌、等。基于多重相似性和增强注意力预测药物-靶标相互作用[J].学报(自然科学版),2025,53(2): 99-107.(Wang Wei,Yu Mengxue,Sun Bin, et al.Prediction of drug-target nteraction based on multiple similarity and enhanced attention mechanisms on graph convolution neural network[J].Journal of Henan Normal University (Natural Science Edition),2025,53(2) :99-107. DOI:10. 16366/j.cnki.1000-2367.2023.06.28.0003.)

基金:

国家自然科学基金;河南省科技攻关项目

关键词:

图卷积神经网络(GCN);多重相似性;PCA;增强注意力机制;药物-靶标相互作用

graph convolution neural network;multiple similarity;PCA;enhanced attention mechanisms;drug-target interaction

分类号:

TP181


基于多重相似性和增强注意力预测药物-靶标相互作用.pdf

Baidu
map