LncRNA-Disease Association Prediction Based On Heterogeneous Network

Long non-coding RNA is a kind of non-coding RNA molecules that longer than 200 nucleotides.More and more studies show that lnc RNAs participate in many biological processes and are associated with complex diseases.Finding lnc RNA-disease association help us understand the process of occurrence and development of the disease and explain its molecular mechanism,so as to further prevent,diagnose and treat diseases.Because of the large number of lnc RNAs and complex diseases,making use of computational methods to predict potential lnc RNA-disease associations can provide direction for biological experiments,so we can save times and money.Therefore,the prediction of lnc RNA-disease association has very important theoretical research value and social value.In view of the limitations of existing lnc RNA-associated disease prediction methods include using circumscribed data and high computing time-consuming on large-scale heterogeneous networks,this paper proposes a lnc RNA-disease association prediction method based on heterogeneous networks.This method integrates data of lnc RNAs,diseases and their associated biomolecules by constructing a heterogeneous network.The integration of existing information to the greatest extent,thereby improving the accuracy.In the prediction process,the concept of meta-path is combined,therefore the type information of each edge is extracted to combined the biological meaning.Add semantic information to the prediction process making the prediction process more reasonable and more interpretable.Using the classification result as a prediction result avoids the iterative calculation process on the entire network that exist in the existing method,which greatly reduces the calculation time.In order to verify the accuracy of the predicted results,we performed a case study by using the predicted result of lnc RNA-disease association which related to liver cancer,gastric cancer and colon cancer.The case study includes four verification aspects,that is whether information was recorded in the database or not,whether there was literature verification information or not,whether expression differences existed in the normal and disease samples or not,and whether they were detected by the other forecasting methods or not.The results show that the accuracy of this prediction method in all three diseases is greater than 99%,and the prediction results are reliable.Since most of the existing prediction methods cannot calculated on such a large and complex heterogeneous network,this paper compares our method with the algorithm based on network propagation proposed by Zhang J et al.The results show that the AUC value of the method in this paper reaches 0.92,which is higher than the method of comparison and the prediction process cost less time.The heterogeneous network-based lnc RNA-disease association prediction method proposed in this paper has good results in both evaluation indicators and case studies,can be applicable to large-scale heterogeneous networks,and also have better flexibility and scalability.Our method have important implication for lnc RNAs-disease association prediction study.