Prediction Of MiRNAs Mature Sequences Through Energy Distribution And Structure

This paper presents a mature sequence prediction of vertebrates microRNAs prediction methods, it was based on the specific miRNA maturation in vivo mechanism, and used energy distribution and structural characteristics. This method has high accuracy, and can provide theoretical guidance for experimental searching for microRNAs.As a research focus in recent years, microRNAs are a class of small non-coding RNAs, it binds the 3'-UTR of mRNAs by specific binding and inhibition of its expression. This class of small RNA moleculars is involved in almost all cellular processes in living organisms, and shows regulatory functions in a variaty of species. microRNA mature sequences is generated from pre-miRNA through Dicer processing, while pre-miRNA is obtained after Drosha cutting pri-miRNA. Northern blotting and cloning have been proven to be efficient in identifying numerous microRNAs in their mature state. However, these types of methods are inevitably biased towards those microRNAs that are abundantly expressed. Recently developed "high-throughput sequencing" is an efficient sequencing technolgy. But such kind of efficiency is also associated with massive data and therefore needs computer mehods for assistance.microRNA's maturation experienced Drosha processing and Dicer processing, and both the substrates form into 'stem-loop structure'. However, due to the phenomena that a large number of genome sequences can form likely 'stem-loop structure', there have to be some code provided by microRNAs for the Drosha and Dicer to recognize its processing site. We currently have several computational methods designed to predict pri-miRNAs by identify mature sequence characteristics, but most of these methods have the short-coming of only marginally addressing this issue of mature microRNA as an inline procedure and only extract information relating to mature microRNAs directly from the sequence of the pri-microRNAs. For example, miRseeker, proMIR. However, as these tools are not specified for mature sequence prediction, their accuracy is low. So far, there is no tools specifically designed for finding microRNA mature sequences directly from pri-miRNAs. Some other methods are able to predict human precursor microRNA sequences (pre-microRNAs), such as Microprocessor SVM. This method utilizes more than 600 features, which makes it impossible to tell the key feature for Drosha to recognize its point.Based on previous statistics on pri-miRNA and pre-miRNA sequences and structures, we designed a new tool MiRmat for vertebrates mature microRNA sequence prediction. Experiment shows MiRmat has relatively high accuracy:for Drosha site 84% while for Dicer site 92.8% when 2nt error is permited.This paper describes a prediction methods, because of its high accuracy, it can be used as theoretical guidance.