MiRNA-Disease Correlation Research Based On Link Prediction

Identifying disease-related microRNAs(miRNAs)is an essential but challenging task in bioin-formatics research.Much effort has been devoted to discovering the underlying associations between microRNAs(miRNAs)and diseases based on link prediction methods.However,most studies mainly focus on designing advanced methods to improve prediction accuracy while ne-glecting to investigate the link predictability of the relationships between miRNAs and diseases.sd In our first work,we construct a heterogeneous network by integrating neighborhood informa-tion in the neural network to predict potential associations between miRNAs and diseases.We also employ a new computational method called a Neural Network model for MiRNA-Disease Association prediction(NNMDA).This model predicts miRNA-disease associations by inte-grating multiple biological data resources.Comparison of our work with other algorithms re-veals the reliable performance of NNMDA.Its average AUC score was 0.937 over 15 diseases in a fivefold cross-validation and AUC of 0.8439 based on leave-one-out-cross-validation.The results indicate that NNMDA could be used in evaluating the accuracy of miRNA-disease as-sociations.Moreover,NNMDA was applied to two common human diseases in two types of case studies.In the first type,26 out of the top 30 predicted miRNAs of lung neoplasms were confirmed by the experiments.In the second type of case study for new diseases without any known miRNAs,we selected breast neoplasms as the test example by hiding the association information between the miRNAs and this disease.The results verified 50 out of the top 50 predicted breast neoplasm-related miRNAs.In addition,based on the neural network-based miRNA disease association prediction(NN-MDA),we conducted our second work.In the second work,we integrated more infomation(miRNA-gene association,disease-gene association,gene-gene association)from some other database,then we reconstructed a more complex heterogeneous network,and the algorithm has been further improved through the combination with multi-task learning.In the five-fold cross-validation,the average AUC value of the 15 diseases reached 0.954.In addition,in two types of case studies,we applied MNNMDA to two common human diseases.In the first type of case s-tudy,46 of the top 50 miRNAs predicted for colonic neoplasms were experimentally confirmed.In the second case study,we selected ovarian neoplasms as test samples by hiding the associ-ations between miRNA and the disease.The results confirmed 48 of the 50 predicted breast neoplasms-associated miRNAs.In addition,we validated the performance of the algorithm in the reconstruction of the gene-disease network.These experimental results are a good proof of the effectiveness and efficiency of our algorithm.