Functional Network-based MicroRNA And Disease Association Prediction Methods

In this paper,the relationship between diseases and miRNA is studied.MiRNA functional similarity calculation,parallel computing of similarity calculation,human disease-miRNA association prediction are studied deeply.Some innovative achievements have been made.This paper mainly includes the following three aspects:(1)The parallel computing method of disease term semantic similarity and miRNA functional similarity is studied.The current calculation methods of disease simantic similarity and miRNA functional similarity take a long time.When studying a new similarity calculation method,the waiting time is difficult to accept after changing the calculation strategy.In order to speed up the calculation of similarity,a parallel computing method for disease term semantic similarity and miRNA functional similarity by using MapReduce framework has been proposed.In contrast to the serial algorithm,we reduce the similarity computation time to an acceptable range,and a single node running the algorithm also has a certain acceleration effect.(2)The calculation method of miRNA functional similarity is studied.At present,miRNA functional similarity calculation method using disease simantic similarity usually calculates the similarity of the two groups of diseases.The inheritance information for the same group of disease terms is calculated repeatly in the process of calculating the information content of a group of disease terms.This paper presents a method for calculating miRNA similarity based on group-wise comparison.We use the method of comparing the information content of the two groups of disease terms to calculate miRNA functional similarity by removing the overlapping information between the disease terms within the groups.Compared with the existing methods,our method has the characteristics of higher similarity with the niRNA functional similarity between the same family and the same cluster of miRNAs,which is in line with the biological premise.In this paper,we compared the performance of three disease-miRNA association prediction method with different miRNA functional similarity methods.Our method shows a higher AUC value in all three prediction methods,(3)The miNRA association prediction method was studied.We use each miRNA as a starting point to independently perform a random walk with restart on a miRNA functional network constructed using miRNA functional similarity to improve prediction performance.Compared with the existing methods,our method has a higher AUC value in the 5-fold cross validation.