Predicting miRNA-disease associations based on graph autoencoders and collaborative training
In recent years,increasing biological experiments have shown that microRNA(miRNA)plays an important role in the development of human complex diseases.Therefore,predicting miRNA-disease associations can contribute to accurate diagnosis and effective treatment of diseases.Since traditional biological experiments are expensive and time-consuming,plenty of computational models based on biological data have been proposed to predict MiRNA-disease associations.In this study,we propose an end-to-end deep learning model to predict miRNA-disease associations(MDAGAC).Specifically,we firstly construct the similarity network of miRNA and disease by integrating disease semantic similarity,miRNA functional similarity and Gaussian interaction profile kernel similarity.Then,the effect of label propagation is improved through Graph Autoencoders and Collaborative training.This model implements two graph autoencoders on miRNA graph and disease graph respectively,and trains these two graph autoencoders collaboratively.Graph autoencoders on miRNA graph and disease graph are able to reconstruct score matrix through initial association matrix,which is equivalent to propagate labels on graphs.The prediction probability of MiRNA-disease association can be obtained from the score matrix.The results of the experiment based on 5-fold cross validation show that MDAGAC is reliable and effective and outperforms current MiRNA-disease associations prediction methods.