Prediction Of Long No-Coding RNA And Disease Associations Based On Multi-Label Learning

Long noncoding RNAs（lnc RNAs）are defined as noncoding RNAs with a length greater than200 nucleotides.Lnc RNAs have been considered as transcriptional noise in the early stage.Until recent studies have proved that lnc RNAs have rich biological functions and participate in almost all life processes.Mutations and disorders of lnc RNAs are associated with many diseases and play an important role in organisms.Therefore,the identification of new lnc RNA-disease will help humans understand the function of lnc RNAs,identify biomarkers and find drugs,and contribute to the diagnosis,treatment,prognosis and prevention of diseases.However,there are still some problems in the existing calculation methods,such as the relatively small association between lnc RNAs and diseases confirmed by experiments,the inherent noise in the data set and the inability to effectively fuse the similarity matrix,which greatly limits the prediction accuracy of the existing models.Therefore,how to reliably and effectively predict disease-specific lnc RNAs remains a challenging task.In order to solve these problems,this thesis mainly does the following research:We propose an approach based on robust multi-label learning to predict lnc RNA-disease associations.Firstly,the algorithm constructs a set of similarity matrices for lnc RNAs and diseases,including disease semantic similarity,lnc RNAs functional similarity,cosine similarity and Gaussian interaction spectrum kernel similarity matrix.Then,the algorithm uses?₁ norm to adaptively update the integrated lnc RNAs and disease similarity matrix to obtain a clearer similarity structure.Finally,the association matrix is updated using a graph based multi marker learning framework to reveal the potential consistency between lnc RNAs space and disease space.The experimental results on five common data sets show that the proposed method can achieve good results in five-fold cross validation and leave-one-out cross validation.The case study of prostate cancer further confirmed the effectiveness of the proposed method in identifying lnc RNAs as potential prognostic biomarkers.We propose an approach based on multi-view and multi-label learning to predict lnc RNA-disease associations.Similarly,the algorithm first constructs a set of similarity matrices for lnc RNAs and diseases.Then,in order to better mine the potential shared information in different similarity views,the algorithm uses the multi-view model guided by consistent graph to obtain the global similarity matrix.In addition,the algorithm integrates the prediction of the association between lnc RNAs space and disease space into a unified framework by using multi marker learning.Experimental results on three commonly used data sets show that the proposed method can achieve good prediction performance in a variety of cross validation scenarios.Finally,the proposed algorithm is applied to the prediction of lnc RNAs related to cervical cancer.Literature mining and survival analysis prove the rationality of the prediction results.