Evaluation of DNA mapping schemes for exon detection

Identification of protein coding regions (exons) in eukaryotic genomic sequences is an active area of research at present. Mapping of symbolic genomic sequences to numeric sequences is the first step required for processing them using digital signal processing (DSP) tools. For DFT-based methods paired numeric and frequency of nucleotide are reported as the best mapping schemes. In this work performance of a wavelet-based method for exon detection is evaluated with different symbolic-to-numeric representations. Optimum performance is obtained by using Z-Curve for DNA mapping. For performance evaluation Receiver Operating Characteristics (ROC) curves are used and the study is conducted on HMR195 data set. This work in general highlights that exon prediction accuracy and computational complexity of the DSP-based algorithms is dependent on the scheme used to map DNA nucleotides into numerical sequences, and so the optimum performance of any algorithm can only be attained with a particular mapping scheme.