基于结构化深度聚类网络的miRNA-疾病关联预测

Prediction of MiRNA-Disease Associations Based on Structural Deep Clustering Network

提出一种结构化深度聚类网络模型来预测microRNA(miRNA)和疾病的关联。模型将miRNA和疾病的集成相似性投入自编码器,将自编码器的输出通过传递算子传递到图卷积层,利用双重监督机制对模型进行训练。5折交叉验证结果显示,该模型分别在HMDD v2.0和HMDD v3.0数据集上平均AUC(Area Under the Curve)值分别为93.23%和94.58%。

A structured deep clustering network model was proposed to predict the association between microRNA(miRNA)and diseases.The model projects the integrated similarity of miRNA and disease into the Autoencoder,and then transfers the new output features obtained from the Autoencoder to the graph convolution layer through the transfer operator.At the same time,the model is supervised and trained using the dual supervision mechanism.The 5-fold cross validation results showed that the model achieved 93.23%and 94.58%average AUC values on the HMDD v2.0 and HMDD v3.0 datasets,respectively.