基于图神经网络和随机森林的CircRNA-疾病预测  

CircRNA-disease prediction based on graph neural networks and random forests

作　　者：王波 尹帅 杜晓昕[1] 张剑飞[1] 周振宇 WANG Bo;YIN Shuai;DU Xiaoxin;ZHANG Jianfei;ZHOU Zhenyu(School of Computer and Control Engineering,Qiqihar University,Qiqihar 161006,China)

机构地区：[1]齐齐哈尔大学计算机与控制工程学院,黑龙江齐齐哈尔161006

出　　处：《高师理科学刊》2024年第2期36-41,47,共7页Journal of Science of Teachers＇College and University

基　　金：2022年度黑龙江省省属高等学校基本科研业务费科研项目(145209125)。

摘　　要：环状RNA(CircRNA)广泛参与人类疾病的进程,其突变和失调与许多人类疾病密切相关.因此,建立一个高效准确的CircRNA与疾病之间的预测算法对于提前对疾病的发生做出预防以及发病后的治疗方案具有重要意义.提出了一种新的基于图神经网络和随机森林的算法预测CircRNA-疾病关联算法,在分层网络表示嵌入部分通过构建异构网络,根据网络图的邻近性,对网络图的节点和边缘进行分层,递归地合并原始图中的节点和边,得到若干具有相似特征的较小子网络.子网络规模随着分层的深入而递减,直至得到最小子网络后,使用node2vec网络图游走算法对其进行预处理,然后将全部节点的特征向量输入至随机森林分类器来识别潜在的CircRNA-疾病关联,从而进行预测.Circular RNA(CircRNA)are widely involved in human disease processes,and their mutations and dysregulation are closely associated with many human diseases.Therefore,establishing an efficient and accurate prediction algorithm between CircRNA and diseases is important for making prevention of disease occurrence in advance as well as treatment programs after the onset of diseases.A new algorithm based on graph neural network and random forest is proposed to predict CircRNA-disease association algorithm,in the hierarchical network representation embedding part by constructing a heterogeneous network,according to the proximity of the network graph,the nodes and edges of the network graph are layered,and the nodes and edges in the original graph are merged recursively to obtain a number of smaller sub-networks with similar characteristics,and the size of the sub-networks decreases with deeper layering until the smallest sub-network is obtained.The size of the sub-networks decreases with the depth of layering until the smallest sub-network is obtained,which is preprocessed using the node2vec network graph wandering algorithm,and then the feature vectors of all the nodes are inputted into the random forest classifier to identify potential CircRNA-disease associations and thus make predictions.

关 键 词：CircRNA-疾病关联预测 图神经网络 node2vec 随机森林 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]