Digital Filter-Based Exon Prediction: Exploring the Reason why Cascading 2nd Order Moving Average Filter Directly with Anti-Notch Filter should be Avoided

In genomic data analysis, DSP plays a crucial role in extracting hidden periodicities and other characteristic features from DNA or protein sequences. An illustrative example of this effective application is observed, where the anti-notch filter (ANF) is employed to extract a three-base periodicity feature from a DNA sequence, which is quite useful in distinguishing exons from introns. However, the effectiveness of ANF techniques is compromised due to the excessive background noise in their period-3 spectrum. This background noise masks the discriminative three-base periodicity (TBP) features, increasing the likelihood of false predictions. To address this concern, various de-noising techniques have been proposed over the past decades. A second-order moving average filter is the most straightforward technique that has also been employed to suppress the background noise. In this work, a crucial aspect of the arrangement of the second-order moving average filter is discussed, which was not addressed in earlier research to the best of my knowledge. It is demonstrated how the direct cascading of a 2nd Order Moving Average Filter with an Anti-Notch Filter diminishes the Three-Base Periodicity (TBP) feature, significantly impacting the overall accuracy of exon prediction.