DefinePK

Abstract

Identification of the regions that code for proteins in a deoxyribonucleic acid (DNA) sequence is a vital and challenging task in the area of Bioinformatics. Study of exon regions is a substantial phenomenon in designing drugs and identification of diseases. The fragments of DNA that contain protein coding information are termed as exons. Hence finding the exon locations in a DNA sequence is a crucial job in genomics. Nucleotides aid as the fundamental structural unit of a DNA. Three base periodicity (TBP) has been observed in the regions of DNA sequences which code for proteins in case of nucleotides. By applying signal processing methods, TBP can be easily determined. Adaptive signal processing methods found to be probable in comparison with several other methods. This is due to the distinctive ability of adaptive algorithms to change weight coefficients depending on genomic sequence. We propose a novel adaptive exon predictor (AEP) based on these deliberations using normalization to improve pursuing ability of the adaptive algorithms. We develop AEPs using LMS algorithm with its data clipped; error clipped and signed normalized variants to reduce computational complexity. Hybrid variants of proposed AEPs include DCLMS, ECLMS, ECLMS, DNLMS, DNDCLMS, DNECLMS, and DNDECLMS algorithms. It was shown that DNDCLMS based AEP is better in exon prediction applications based on performance measures with Sensitivity 0.6872, Specificity 0.7043 and precision 0.6722 at a threshold of 0.8. Finally the capability of several AEPs in predicting exon locations is verified using different genomic sequences found from National Center for Biotechnology Information (NCBI) database.