Development And Improvement Of The Microrna Classification Algorithm Based On Pattern Recognition

With the constant development of biology, more and more advanced technologies have been applied into biological research. In this paper, Artificial Neutral Networks, Support Vector Machines (SVM) and other computer pattern recognition methods are combined with biology, which helps better solve the biological problems. As one tiny non-coding RNA, miRNA distributes in the eukaryote with a length of 21～25 nucleotides. There were only one or two small molecules in the initial study. But then a great number of miRNAs in different species were found, which have been emphasized on the adjustment function.At present, the methods of identifying miRNA can approximately be classified into the biological experiment and computer method. With higher costs and longer experimental cycle, the biological experiment accuracy is verified, which constraints its large-scale application. Due to low cost, high efficiency, handling of massive information, suitable for large-scale forecast and other advantages, the computer method has developed rapidly. According to different recognition algorithms, the method can be classified into two categories: one is based on the comparative genomics, the other is based on the artificial intelligence methods. The former method needs the miRNA information which has a high accuracy in the same resource sequence with the control. The later one, named as ab initio prediction method, can forecast the recognition sequence independently.Due to the effect of feature extraction on the accuracy of recognition algorithm, to find a suitable feature is a key to design intelligence recognition algorithm. Based on the secondary structure of pre-miRNA and former work, the bilateral synchronous sliding window, which focuses on the collection and statistics of useful feature, is introduced in this paper. In this method, the redundant information is deleted, and the dimensions of feature sector are reduced, which contributes to high efficiency.The Artificial Neural Network and SVM are widely applied in the field of pattern recognition. Based on the artificial neural network algorithms and bilateral synchronous sliding window method, micro-RNA identification procedures are developed in this paper. And the high accuracy and efficiency are verified by identifying well-known human micro-RNA sequence. And the features collection and statistics are also verified by using SVM and bilateral synchronous sliding window method.