Improved Exon Prediction Technique by De-noising Period-3 spectrum with SVD Algorithm

The rapid growth of DNA sequencing technology has made it more convenient for researchers to study hidden patterns and periodicities of genomic signals which help in determining different functional properties of DNA structures. One such key periodicity is three base periodicity (period-3 behavior) which is mostly observed in protein coding regions of DNA signals and therefore it is widely used in protein coding region finding applications. Spectral analysis using narrowband pass filters has already been used for this purpose but the existing background noise of DNA signals limit the performance of these techniques. So far, a number of de-noising techniques have been used to reduce the effect of this noise. This paper emphasizes the effectiveness of singular value decomposition (SVD) as a de-noising filter for removal of background noise. The results are compared with the conventional moving average filter. The results obtained by SVD method demonstrate that the SVD method gives better de-noising performance than moving average filter and hence improve the prediction accuracy of exon prediction significantly.