基于异构图神经网络的微生物与疾病关联挖掘

Microbe-disease Association Mining in Heterogeneous Graph Neural Network

微生物与疾病关联挖掘的工作大多未考虑微生物分类信息树等微生物背景信息,也未能充分利用微生物与疾病间的非线性关联,为此提出一种基于异构图神经网络的微生物与疾病关联挖掘方法。利用微生物分类信息树完善生物分子网络,输出异构生物分子网络节点的低维向量并挖掘节点间关联。研究结果表明,与传统方法相比,引入微生物分类信息树的新方法具有更好的效果,AUC、AUPRC和F1-score分别达到0.901、0.552和0.557。

Most of the existing work on microbes-diseases association mining fails to take into account the microbial background information,such as microbial taxonomy hierarchy tree,and fails to make full use of the nonlinear association between microbes and diseases.There fore a method of association mining between microbes and diseases in heterogeneous graph neural network is proposed.The method perfects the biomolecular network through microbial taxonomy hierarchy tree,output the low dimensional vector and mining association between microbes and diseases.The results show that the new method has better results compared to the traditional method,with AUC,AUPRC and F1-Score up to 0.901,0.552,0.557.