Research And Implementation Of Predicting Human Disease-related MicroRNAs

MicroRNAs(miRNAs) are a set of endogenous non-coding RNAs with lengthof about22nt. Based on the complementary of sequence pair, They regulate theexpression of the target mRNA, thereby controlling gene expression. Theabnormal expression of genes is an important factor for a variety of disasese. SomiRNAs play an indirect role in the accurence of diseases. This paper proposed andimplemented methods of predicting disease-related miRNAs. These methodsachieved higher prediction performance.First, according to what National Library of Medicine define a disease, wecalculated the similarity of every two diseases. And then we use this similariy tocalculate the similarity of the two miRNAs with the association data of HMDD. Weconstructed the miRNA function similarity network, using miRNAs as the nodes,the similarity as the weight of edges.Secondly, according to that the more similar miRNAs are, the more possibllythey regulate the same disease, we proposed a prediction algorithm based on K-nearest neighbors model. We also proposed weight-added prediction methodconsidering the information of families and clusters. Otherwise, we proposed amethod of predicting disease-related miRNAs based on a random walk model onthe whole miRNA function similiarity network. These algorithms have better resultcomparing with existing miRNA-disease association prediction methods.Finally, we also established a website which provides the predicted result of18kinds of common human diseases. It also has the functions of searching the knownmiRNA-disease associations and analyzing the relation of multi-miRNAs which canhelp the biologists online services of searching and analyzing.