GC Wave Analysis in Promoter Regions via Wavelet Analysis and Support Vector Machine

Abstract

A large number of genomes have been sequenced and the number is growing rapidly. It is crucial to improve sequence annotation, including promoter prediction. Many aspects of DNA sequences have been examined and used in promoter prediction. In particular, the physical instability correlating GC content in the promoter region has been focus of many studies. To extract the GC signals of a promoter region in a genome sequence, we adopt a scheme combining wavelet analysis and a support vector machine (SVM). In this scheme, we take a simplified way to quantize and extract chemo-physical properties of a DNA sequence. Four types of DNA are converted to binary form with respect to G and C or not. The sequences are expanded to two dimensional spaces, frequency and location, by discrete wavelet transformation (DWT). The fixed length of the promoter and randomly selected DNA segments are prepared as the positive and negative training data, respectively. The two types of data are converted by DWT and learned by a SVM. Then, previously unknown DNA segments are classified as promoter or nonpromoter by the trained SVM.