Research On Method Of Computing The Functional Similarity Of Human MiRNAs Based On Targets

MicroRNAs(miRNAs)are a major class of endogenous small non-coding RNA molecules that regulate gene expression by binding 3' untranslated regions of the target mRNAs at the post transcription level.MiRNAs participate in many critical biological processes,and therefore,miRNAs are considered to be related with various diseases including cancers.Revealing the function of miRNAs can help us develop new drugs for the diagnosis and treatment of diseases.Although thousands of miRNAs have been identified up till now,the functions of most miRNAs are still unknown.The development of high-throughput technology has accumulated large amounts of available data,providing the possibility of computing the functional similarity of miRNAs and inferring the functions of miRNAs.Due to the shortage of existing methods,we propose two methods for quantifying the functional similarity of human miRNAs based on targets.For protein coding genes,the functional similarity among them can be computed based on their sequence,expression similarities or Gene Ontology(GO)annotations.The methods based on GO for measuring functional similarity of protein coding genes always achieve better results.With the rapid accumulation of GO annotations for genes,it is feasible to measure functional similarity of miRNAs based on GO annotations of their target genes.In this paper,we propose a novel method,called MFS_GO,to quantify the functional similarity of human miRNAs based on GO annotations of the targets.The comparison results show that MFS GO outperforms miRFunSim on reflecting the correlation between functional similarity and expression similarity of human miRNAs as well as the relationship between miRNAs and diseases.Furthermore,MFS_GO is applied to a case study of breast cancer and the result indicates the validity of the method that MFS_GO can effectively uncover novel candidate breast cancer-related miRNAs.The functional associations among targets can be reflected by the corresponding products,which also can be reflected by Protein-Protein Interaction(PPI)network.Considering both aspects mentioned above,we propose a novel method,named MFS_NC,to compute the functional similarity of human miRNAs based on network connectivity.Several experiments indicate that MFS_NC has better performance than miRFunSim and MFS_GO on reflecting the correlation between functional similarity and expression similarity of human miRNAs as well as the relationship between miRNAs and diseases.MFS_NC can effectively identify novel candidate breast cancer-related miRNAs.In addition,MFS_NC performs slightly better than MFS_GO on identifying novel candidate breast cancer-related miRNAs.