56 Promoter Recognition with Wavelets and an SVM

Promoter recognition by computer contributes to improving gene annotation, including gene identification and prediction. We propose an alternative approach for promoter recognition using wavelets and a support vector machine (SVM) to introduce a new perspective for promoter recognition. A wavelet transformation, similar to a Fourier transformation, maps a function onto a two-dimensional space; frequency and location. SVMs are a set of supervised learning methods used for classification and regression. A set of promoter DNA segments and non-promoter segments are used as positive and negative data in our method. A DNA segment is translated numerically and expanded to a two-dimensional space by a wavelet transformation. The transformed positive and negative data are used as training data for the SVM to discriminate between promoters and non-promoters. Then, unknown DNA segments are classified as promoters or non-promoters with the trained SVM.