Detects conserved miRNAs. miRRim2 can not only accurately detect miRNA hairpins, but infer the location of a mature miRNA sequence. In miRRim2, each position of a miRNA hairpin is expressed as a multidimensional feature vector to detect position-specific features; therefore, a miRNA hairpin is expressed as a sequence of the feature vectors. miRNA hairpins, expressed by sequences of feature vectors, are modeled using conditional random fields (CRFs, which optimize feature weights so that a trained model can most probably discriminate between miRNA hairpins and background data. The probabilistic model used in miRRim2 has several sub-components, each of which corresponds to a specific component of miRNA hairpins, such as mature miRNA, passenger strand, and terminal loop regions; therefore, the position-specific features of each component are appropriately modeled.
INTEC Inc., Tokyo, Japan; Computational Biology Research Center, National Institute of Advanced Industrial Science and Technology (AIST), Tokyo, Japan; Department of Computational Biology, Graduate School of Frontier Sciences, University of Tokyo, Chiba, Japan
miRRim funding source(s)
This work was supported partly by a Grant-in-Aid for Scientific Research (A) and the Functional RNA Project of the New Energy and Industrial Technology Development Organization.