A computational approach for prediction of exons using static encoding methods, digital filter and windowing technique

Introduction Identifying protein-coding regions in eukaryotic Deoxyribonucleic acid (DNA) remains difficult due to the sparse and uneven distribution of exons. Methods This work focusses into four static encoding schemes—integer, Voss, paired numeric, and Electron-Ion Interaction Potential (EIIP) to improve exon prediction using genomic signal processing. Two benchmark sequences, Caenorhabditis elegans Cosmid F56F11.4 and Mouse apolipoprotein A-IV (M13966.1), were analyzed in MATLAB. A Cauer (elliptic) band-pass filter was used to isolate the period-3 component, and a Blackman-Harris window was utilised to reduce spectral leakage. The elliptic filter in conjunction with EIIP-based encoding achieved the most distinct separation between coding and non-coding areas among the assessed techniques, identifying every exon segment with a minimal amount of noise. Results and discussion The technique obtained 84% sensitivity, 96% specificity, and 94% accuracy on the C. elegans Cosmid sequence and 86.5% sensitivity, 93% specificity, and 91% accuracy on the M13966.1 gene sequence. Conclusion These results show that the EIIP, Cauer filter and Blackman-Harris windowing framework offers a reliable and effective method for identifying exons.