Long Non-coding RNA-Disease Association Prediction Based On Network Consistency Projection

In recent years,long non-coding RNA(lncRNA)has attracted attention because of its emergence in many important biological mechanisms.More and more biological experiments and clinical reports indicate that lncRNA has multiple functions as tumor suppressors,oncogenes,transcriptional regulators and epigenetic modifiers.lncRNA is closely related to the occurrence and development of various complex and serious diseases in human beings,and studying the potential relationship between lncRNA and diseases has become an important topic for many researchers.However,only a few lncRNA and disease have been experimentally verified,so it is very urgent to study the potential relationship between lncRNA and disease.This not only helps to elucidate the complex pathogenesis,but also helps the diagnosis,treatment and prognosis of the disease.However,understanding the relationship between lncRNA and disease through biological experiments and clinical research requires a lot of human and material costs.However,understanding the relationship between lncRNA and disease through biological experiments and clinical research requires a lot of human and material costs.At present,Big data and artificial intelligence technologies are rapidly emerging more and more researchers apply computer technology and statistical methods to bioinformation data mining.Many researchers have proposed calculation methods to predict potential lncRNA-disease associations.And many prediction models have achieved considerable prediction effects.Although these prediction models have achieved good prediction results,most models still have a lot of room for improvement.On the one hand,this type of research requires high data richness.Many models use less lncRNA and disease similarity.On the other hand,it is impossible to predict lncRNA without known related diseases.In order to solve these problems,this article did the following research:(1)We have integrated lncRNA expression similarity network,lncRNA cosine similarity network,disease semantic similarity network and disease cosine similarity network.(2)Based on the assumption that functionally similar lncRNAs tend to be associated with diseases with similar phenotypes,and vice versa.We propose a new predictive calculation model,the lncRNA-disease association prediction model NCPHLDA based on the consistency of network projections.The core method of NCPHLDA prediction model is not only a parameterless method,but also does not require negative samples.NCPHLDA prediction model can predict new lncRNA(there is no known diseases associated with it).(3)In terms of model evaluation,we divide performance evaluation and case studies,and use cross-validation methods for performance evaluation.By implementing leaveone-out cross-validation and five-fold cross-validation on NCPHLDA prediction model,we can obtain AUC values of 0.9273 and 0.9179 ± 0.0043.In comparison with some classic prediction models,we also adopted leave-one-out cross-validation and five-fold cross-validation for comparison.The results show that NCPHLDA prediction model has better prediction capabilities.In addition,the performance comparison of cosine similarity and Gaussian kernel interaction similarity also shows that cosine similarity performance is better.In case studies,case studies were conducted on three diseases(breast cancer,cervical cancer and hepatocellular carcinoma),and the results showed that NCPHDLA prediction model had reliable and convincing prediction performance in predicting potential human lncRNA-disease association.