Recognition of Alternative Splice Sites Based on Discrete Cosine Transform Learned by SVM with the Mixed Pseudo-Image Coding Method

With the development of bioinformatics and the nearing completion of the Human Genome Project, the task of optimizing the methods of gene sequencing is becoming more and more significant. Alternative splicing, which plays an important role in regulating gene expression, widespreads in eukaryote genes. So developing a better analysis algorithm for the recognition of alternative splicing sites is a matter of great urgency. In this paper, we developed an accurate alternative splice sites identification algorithm based on discrete cosine transform (DCT) learned by support vector machines (SVM) using maximum-margin principle. And we also developed a new pseudo-image coding method mixed with one-dimensional coding method and two-dimensional coding method. Experiment results demonstrate that this new approach can achieve better or similar performance for alternative splice sites recognition than other methods.