A new classification method for human gene splice site prediction

Abstract

Human splicing site prediction is important for identifying the complete structure of genes in Human genomes. Machine learning method is capable of distinguishing the different splice sites in genes. For machine learning method, feature extraction is a key step in dealing with the problem of splicing site identification. Encoding schema is a widely used method to encode gene sequences by feature vectors. However, this method ignores the information of the period-3 behavior of the splice sites and sequential information in the sequence. In this paper, a new feature extraction method, based on orthogonal encoding, codon usage and the sequential information, is proposed to map splice site sequences into feature vectors. Classification is performed using a Support Vector Machine (SVM) method. The experimental results show that the new method can predict human splice sites with high classification accuracy. © 2012 Springer-Verlag.