Studies of spectral properties of short genes using the wavelet subspace Hilbert-Huang transform (WSHHT)

This paper presents a new algorithm for the analysis of spectral properties of short genes using the wavelet transform and the Hilbert-Huang transform (HHT). A wavelet subspace algorithm combined with the empirical mode decomposition (EMD) is introduced to create subdivided intrinsic mode functions (IMFs) and a cross-correlation analysis is applied to remove pseudo-spectral components. Experiments are carried out on DNA sequences with the double-base (DB) curve representation and the results show that the signal-to-noise ratio of buried signals can be enhanced using the proposed method, yielding significant patterns that are rarely observed with conventional methods. The wavelet subspace Hilbert-Huang transform (WSHHT) algorithm is able to correctly identify spectral patterns of very short genes (below 70 bp) in DNA sequences. © 2008 Elsevier B.V. All rights reserved.