Research On LncRNA-disease Association Prediction Based On Matrix Completion

In recent years,more and more biological experiments and clinical reports have shown that long non-coding RNA(lncRNA)occupies a large proportion of biological RNA and plays a key role in various biological processes,including cell growth and development,gene expression regulation,alternative splicing and nuclear organization.Abnormal expression or dysfunction of lncRNA is closely related to various complex human diseases.Therefore,predicting the association of lncRNA-disease is necessary for the diagnosis and treatment of diseases.There are a large number of lncRNA and complex human diseases,and biological experiments to verify the correlation between lncRNA-disease are time-consuming and expensive,so the design of effective computational models is of great significance to the development of bioinformatics.In response to the above problems,the main tasks completed in this article are as follows:1.Regarding the problem of lncRNA-disease association prediction as a recommendation system problem,this paper proposes a lncRNA-disease association prediction method with a complete matrix(MCLDA).MCLDA first calculates the similarity of two different types of lncRNA and disease and performs network fusion,principal component analysis was used to extract the main feature vector matrix of lncRNA and disease similarity.For a new lncRNA,the lncRNA-disease adjacency matrix is rewritten according to the interaction of its neighbors to solve the cold start problem.Finally,the correlation prediction is completed based on the matrix complete model framework.2.Considering that lncRNA with similar functions contain similar association patterns,and diseases with similar phenotypes have similar association patterns.They are the internal structure of lncRNA and disease in the association adjacency matrix.This paper proposes a lncRNA-disease association prediction method based on regularized matrix completion(MCRLDA).MCRLDA uses manifold learning,treats the internal information of the matrix as a regularization item introduced into the basic structure,and combines the approximation of the lncRNA association pattern and the disease association pattern to restore and complete the association matrix.This paper proposes the lncRNA-disease association prediction method based on complete matrix.The experimental results show that the prediction performance of MCLDA and MCRLDA is better than some previous public prediction methods.In order to further verify the performance of the model,this article analyzes related cases of several major human diseases...