Support vector machine for prediction of meiotic recombination hotspots and coldspots in Saccharomyces cerevisiae

A method for predicting hotspots and coldspots using support vector machine (SVM) based on statistical learning theory is developed. This method is applied to published 303 hot and 48 cold open reading frames (ORFs) in Saccharomyces cerevisiae. The sequence features of general dinucleotide abundance and dinucleotide abundance based on codon usage are extracted, and then the data sets are classified with different parameters and kernel functions combined with the method of two-fold cross validation. The result indicates that 87.47% accuracy can be reached when classifying hot and cold ORF sequences with the kernel of radial basis function combined with dinucleotide abundance based on codon usage.