A Research On Disease-related Non-coding RNA Prediction Model Based On Complex Network

Accumulating studies have shown that non-codingRNAs(ncRNAs)are involved in many biological processes and play important roles in a variety of complex human diseases.Studying the relationship between diseases and ncRNAs can not only help us understand the disease pathogenic mechanisms at the lncRNA molecular level,but also promote the diagnosis,treatment,prognosis,and prevention of human diseases.But traditional methods of exploiting potential ncRNA-disease relationships through biological experiments are time-consuming and labour-intensive.Therefore,it is important to develop an effective computational model to identify potential relationships between ncRNAs and diseases.In this paper,we use the known lncRNA-disease associations and miRNA-disease associations to construct complex network to predict the potential associations between two kinds of ncRNAs,long non-codingRNA(lncRNA)and microbeRNA(miRNA)and diseases,respectively.The main research content is as follows:For lncRNA-disease association prediction: we model the prediction problem of potential lncRNA-disease association as the problem of resource allocation on the lncRNA–disease bipartite network and proposed a predictive model named NBLDA.First,two novel lncRNA-disease weighted networks were constructed based on known lncRNA–disease associations and topological similarity of the lncRNA–disease association network,and then the lncRNA-lncRNA weighted matrix and the diseasedisease weighted matrix were obtained based on resource allocation strategy of unequal allocation and unbiased consistence.Finally,label propagation algorithm is used to calculate the potential lncRNA-disease association probability.The results of leaveone-out cross validation(LOOCV)and case studies show that the proposed method is significantly superior to the existing classical algorithm models of the same type.For miRNA-disease association prediction: based on the definition of bipartite network common neighbors and the theory of local community paradigm(LCP-theory),we propose a computational model NBMDA for predicting potential miRNA-disease associations.Considering that the known miRNA-disease associations are very rare,and many diseases(or miRNAs)are associated with only one or a few miRNAs(or diseases),the K-nearest neighbor algorithm(KNN)is utilized as a recommendation algorithm to solve the problem of known association sparsity to improve its prediction accuracy based on known miRNA-disease associations,miRNA functional similarities,diseases semantic similarity and Gaussian kernel similarity for miRNAs and diseases.The results of LOOCV and case studies show that the proposed method is significantly better than the existing classical algorithm model of the same type.