Computational prediction of MicroRNA-Cancer relations

Every year over a million people lose their lives because of cancer, which is why scientists are moremotivated than ever to carry out numerous researches to understand the fundamental causes of cancer ۱.MicroRNAs (miRNAs) are short RNA molecules that bind mRNA, altering their regulation ۲. It is feasibleto use them as potential diagnostic biomarkers ۳ due to their stability and aberrant expression in thepathogenesis of cancer, cardiac, and other diseases ۴. They modulate many processes that contribute todifferent stages of cancers ۵. For example, Experimental evidence suggested that MIR۲۱ could be involvedin different cancers ۶. Many databases provide detailed information regarding these small molecules ۷. Linkprediction is utilized to estimate hidden relations among existing links based on known topology ۸, which ismore efficient in a complex network to find missing links or even predict other ones ۹.Furthermore, it promotes discovering the new relations among disease-related miRNAs that could broadenour horizons of the molecular mechanisms of human diseases ۱۰, which could sometimes be challenging inbioinformatics research ۱۱. The accuracy of new algorithms in link prediction has been proved by extensiveexperiments on both synthetic and real networks ۷. In this article, we have created a miRNA-cancer networkusing four link prediction algorithms, including common neighbors (CN), Preferential attachment (PA),Jaccard (JC), and Adamic and Adar (AA), based on the Human microRNA Disease Database (HMDD v۳.۲)to predict new miRNA-cancer relations. According to our predictions, hsa-mir-۱۴۶a-Glioma and hsa-mir-۱۹a-Non-small-cell lung carcinoma (NSCLC), are some of the most probable Hsa-miRNA cancerassociations that have not been published in any articles yet, and they could be the best probable candidatesfor additional laboratory and validation studies.

Keywords:

miRNA , Link prediction , Cancer , bipartite network , bioinformatics