Frequency-domain analysis of biomolecular sequences

Abstract

Motivation: Frequency-domain analysis of biomolecular sequences is hindered by their representation as strings of characters. If numerical values are assigned to each of these characters, then the resulting numerical sequences are readily amenable to digital signal processing.

Results: We introduce new computational and visual tools for biomolecular sequences analysis. In particular, we provide an optimization procedure improving upon traditional Fourier analysis performance in distinguishing coding from noncoding regions in DNA sequences. We also show that the phase of a properly defined Fourier transform is a powerful predictor of the reading frame of protein coding regions. Resulting color maps help in visually identifying not only the existence of protein coding areas for both DNA strands, but also the coding direction and the reading frame for each of the exons. Furthermore, we demonstrate that color spectrograms can visually provide, in the form of local 'texture', significant information about biomolecular sequences, thus facilitating understanding of local nature, structure and function.