Research On LncRNA-disease Association Prediction Method Based On Biological Networ

Long non-coding RNA(lnc RNA)is a type of RNA with a single-stranded structure and over200 nucleotides,which does not encode proteins and was once considered as noise in the transcription process.In recent years,numerous studies have shown that lnc RNAs play important roles in the formation and development of many human diseases.Investigating the association between lnc RNAs and human diseases is of great significance for promoting disease treatment and the development of related drugs.However,there are a variety of lnc RNAs and diseases,and exploring the associations between different lnc RNAs and different diseases in the laboratory is a time-consuming and labor-intensive task.Therefore,computational methods have obvious advantages and have gradually become a hot research field.Currently,many computational prediction methods have been developed both domestically and internationally,and have achieved good prediction results.However,there are still some problems with existing methods: the biological networks constructed by many methods are too simple to accurately describe the associations between biological elements;the existing lnc RNA-disease association information is too limited,resulting in a sparse known association matrix that limits the improvement of prediction accuracy;at the same time,these methods ignore the influence of other biological elements outside lnc RNAs and diseases.To overcome these problems,this paper proposes three new lnc RNA-disease association prediction methods,which achieve better prediction results.The specific research content is as follows:(1)A matrix completion method based on random walk,BRWMC,is proposed.First,BRWMC constructs multiple lnc RNA(disease)similarity networks based on different measurement perspectives,and uses a similarity network fusion method to merge multiple similarity networks into an integrated similarity network.In addition,the known lnc RNA-disease association matrix is preprocessed using a random walk method,and the estimated scores of potential lnc RNA-disease associations are calculated.Finally,the matrix completion method is used to accurately predict potential lnc RNA-disease associations.(2)A label propagation-based restart bi-random walk method,LPBRW,is proposed.Firstly,the comprehensive similarity network of lnc RNA(disease)is calculated using the similarity kernel fusion method based on lnc RNA functional similarity and Q kernel similarity.In order to solve the problem that the traditional random walk method cannot predict lnc RNA(disease)without any known associations,the label propagation method is introduced.Lnc RNA(disease)without any known associations can learn the association information of adjacent nodes through label propagation,so that they can also participate in prediction and improve prediction accuracy.(3)A method called MANBNR is proposed based on a multi-layered association network and bipartite network recommendation algorithm.Firstly,MANBNR introduces the kernel neighborhood similarity to measure the similarity between different lnc RNAs(diseases).Specifically,MANBNR constructs a lnc RNA-mi RNA-disease multi-layered association network by integrating the lnc RNA-mi RNA association network,mi RNA-disease association network,and mi RNA-mi RNA similarity network,and develops a new method to extract lnc RNA-disease association information from this multi-layered association network.Finally,based on this multilayered association network,the bipartite network recommendation algorithm is used to predict potential lnc RNA-disease associations.The three methods proposed in this paper are all based on biological networks for predicting lnc RNA-disease associations,and they respectively improve the problems of unclear description of biological networks,sparse known lnc RNA-disease association matrices,and simplistic biological elements.The three methods proposed in this paper have achieved excellent results on multiple evaluation criteria,contributing to the field of lnc RNA-disease association prediction and providing new ideas for the treatment of diseases.