Visualization of the protein-coding regions with a self adaptive spectral rotation approach

Abstract

Identifying protein-coding regions in DNA sequences is an active issue in computational biology. In this study, we present a self adaptive spectral rotation (SASR) approach, which visualizes coding regions in DNA sequences, based on investigation of the Triplet Periodicity property, without any preceding training process. It is proposed to help with the rough coding regions prediction when there is no extra information for the training required by other outstanding methods. In this approach, at each position in the DNA sequence, a Fourier spectrum is calculated from the posterior subsequence. Following the spectrums, a random walk in complex plane is generated as the SASR's graphic output. Applications of the SASR on real DNA data show that patterns in the graphic output reveal locations of the coding regions and the frame shifts between them: arcs indicate coding regions, stable points indicate non-coding regions and corners’ shapes reveal frame shifts. Tests on genomic data set from Saccharomyces Cerevisiae reveal that the graphic patterns for coding and non-coding regions differ to a great extent, so that the coding regions can be visually distinguished. Meanwhile, a time cost test shows that the SASR can be easily implemented with the computational complexity of O(N).

An improved random-walk approach for visualizing periodicities hidden in biological sequences

“…The TPM is a 4×3 matrix, each row i (i = 1, 2, 3, 4) stands for a nucleotide base Λ (Λ = A, T, C, or G), each column stands for a "period position" j (j = 1, 2, 3), and the entry m ij (or m Λj ) is the count by which the base Λ appears at the period position j. As a previous work [24], in the original SASR, the TPM of the posterior subsequence at each position t is calculated recursively from t = N to 1, with the recurrence formula and initial value:…”

“…2, for instance). These walk patterns are clues to the discrimination between TP-intensive and non-TP regions [24]. Moreover, the walk shifts in direction from a TP-intensive region to a neighboring one, and the angle of the "corner" indicate the frame shift Δ between the two regions, following a "corner rule" [24].…”

“…These walk patterns are clues to the discrimination between TP-intensive and non-TP regions [24]. Moreover, the walk shifts in direction from a TP-intensive region to a neighboring one, and the angle of the "corner" indicate the frame shift Δ between the two regions, following a "corner rule" [24]. According to Chen and Ji [24], the TP walks for TPintensive sequences (TIS) without any non-TP region have an obvious trend to move rightward and those for non-TP sequences (NTS) without any TP-intensive region move randomly around the zero point.…”

“…It was first designed for the 3-bp periodicity (TP) in DNA sequences, and employed for training-free protein-coding region prediction [24,25], but its principle could be extended for all kinds of periodicities. The SASR takes the query sequence as its input and output a random walk in the complex plane, called the TP walk, which conveniently presents the locations of TP intensive regions as well as frame shifts.…”

“…In Chen and Ji's work [24], they claimed that for non-TP sequences (NTS) without any TP intensive region, the TP walks should be random around zero point, and for TPintensive sequences (TIS) without any non-TP region, the walks should obviously move rightward (details about the original SASR in section 2). However, in this work (see section 3), a slightly leftward trend (an abnormity) in TP walks is observed, which is not reported in the previous SASR literatures.…”