Measuring Function Similarity Of MiRNA By Using Gene Ontology

The function similarity of miRNAs can be applied to predict the function of unknown miRNAs and study the interaction of miRNAs. Sequence and structure similarity of miRNA can be applied to measure their function similarity. However, some miRNAs have similar sequence and structure while have distinct functions. Thus, an effective method should be proposed to measure the function similarity of miRNAs. The function similarity of target genes can be computed by Gene Ontology (GO).This paper introduces a hybrid value of gene GO term to measure the similarity between two sets of GO graphs corresponding to the two sets of target genes of two miRNAs. The spectral clustering algorithm is applied to cluster miRNAs to verify the feasibility of this method. This similarity is then transferred to be the similarity between the two miRNAs. The definition of our information content is mainly based on the size of the term’s descendants, and adjusted by a weight derived from the depth and the GO relationships in the path to the root node. A self-tuning technique and a variance measurement between inter clusters are applied to determine optimal parameters for the spectral clustering of the adjacency matrix of the miRNAs and the number of cluster.The results demonstrate that this method is not sensitive with the time complexity and can cluster the similar function of miRNAs into the same group. It has better clustering performance than the conventional edge-based or node-based method. This method has also demonstrated to be useful for the function annotation of new miRNAs.