An upstream interacting context based framework for the computational inference of microRNA functions.

With the rapid accumulation of microRNA (miRNAs), a class of newly identified small noncoding RNAs, in silico inference of miRNA functions has become one of the central tasks in miRNA bioinformatics. Traditional methods have helped in the understanding of miRNAs, but they also have limitations. In this paper, we first gave a brief review for the progress of bioinformatic methods in miRNA function inference and next presented a new framework (miRUPnet) for inferring the functions of miRNAs by functional analysis of a novel dimension of miRNA network, the context of its transcription factors (TFs) in a protein-protein interaction network. This dimension represents specific biological processes initiated by TF combinations and therefore differs from traditional methods in concept. To validate the accuracy of our method, we first comprehensively mined literature-reported miRNA functions and then made a comparison with the prediction result. The results show that even using the stringent TFBS rule, our method has independently predicted 68.2% of the literature reported miRNA functions, suggesting that miRUPnet has a high accuracy. Moreover, our approach successfully predicted specific functions that could not be inferred for given miRNAs using traditional methods. More importantly, it can distinguish miRNAs from the same family, as well as those present in multiple copies that cannot be differentiated through traditional methods. This study presents a new concept and dimension for miRNA function inference. miRUPnet represents an important and novel method for inferring the function of miRNAs. miRUPnet is available at http://cmbi.bjmu.edu.cn/mirupnet.