Research On The Prediction Method Of MiRNA And Disease Association Based On Deep Learning And Network Representation Learning

Identification of disease-associated mi RNAs(disease mi RNAs)are critical for understanding etiology and pathogenesis.Most previous methods focus on integrating similarities and associating information contained in heterogeneous mi RNA-disease network.However,these methods establish only shallow prediction models that fail to capture complex relationships among mi RNA similarities,disease similarities,and mi RNA-disease associations.(1)A prediction method of mi RNA-disease association based on convolutional neural network and matrix factorizationWe propose a prediction method,CNNMDA,based on network representation learning and dual convolutional neural networks to predict disease mi RNAs.CNNMDA deeply integrates the similarity information of mi RNAs and diseases,mi RNA-disease associations,and representations of mi RNAs and diseases in low-dimensional feature space.The new framework based on deep learning was built to learn the original and global representation of a mi RNA-disease pair.First,diverse biological premises about mi RNAs and diseases were combined to construct the embedding layer in the left part of the framework,from a biological perspective.Second,the various connection edges in the mi RNA-disease network such as similarity and association connections were dependent on each other.Therefore,it was necessary to learn the low-dimensional representations of the mi RNA and disease nodes based on the entire network.The right part of the framework learnt the low-dimensional representation of each mi RNA and disease node based on non-negative matrix factorization,and these representations were used to establish the corresponding embedding layer.Finally,the left and right embedding layers went through convolutional modules,to deeply learn the complex and non-linear relationships among the similarities and associations between mi RNAs and diseases.Experimental results based on cross validation indicated that CNNMDA yields superior performance compared to several state-of-the-art methods.Furthermore,case studies on lung,breast,and pancreatic neoplasms demonstrated the powerful ability of CNNMDA to discover potential disease mi RNAs.(2)A prediction method of mi RNA-disease association based on convolutional neural network and autoencoderMatrix factorization was a traditional dimensionality reduction method,through iterative fitting original similarity and correlation matrix to obtain the characteristic information,the way was very difficult to capture the mi RNA and disease in the lower dimensional characteristics of nonlinear information.Owing to the traditional matrix decomposition method is difficult to capture the mi RNA and diseases of the nonlinear characteristics of information,we put forward a kind of deep web learning method,through the autoencoder for each said low-dimensional mi RNAs and disease characteristics,which in the low-dimensional features of the autoencoder was based on the original mi RNAs and disease characteristics of convolution operation,in order to capture the nonlinear characteristic information between mi RNA and disease,to get a global representation of mi RNA and disease.We built a model based on dual convolutional neural network to learn the original and global representation.The left part was still the embedding of the original mi RNAs-disease node pair,while the right part first obtains the low-dimensional feature representation of mi RNA and disease through the autoencoder,so as to obtain the low-dimensional feature node pair of mi RNAs-disease.Finally,convolutional neural network is used to capture the complex and nonlinear characteristics of left and right paths respectively.Finally,through the case analysis of common diseases to confirm the prediction results.