An Improved Fourier Method for DNA Sequence Classification

The theory and methods of signal processing are becoming increasingly important in bioinformatics and systems biology. The ordinary Fourier analysis is satisfactory for the long DNA sequences to detect period-3 property, but is without impressive success for the short DNA sequences. An improved Fourier method is proposed to increase the accuracy of gene identification by amplifying period-3 behavior in protein coding regions. The experimental results indicate that the improved Fourier approach is better than the ordinary Fourier method and the lengthen-shuffle Fourier algorithm. For the different length window, the accuracy (average of the sensitivity and the specificity) of the improved method is about 3% higher than that of the lengthen-shuffle Fourier method and 9% higher than that of the ordinary Fourier method.