Efficient Transmembrane Segment Prediction in Transmembrane Proteins Using Wavelet-Based Energy Criteria

Integral membrane proteins comprise a class of proteins of special interest. Despite their biological and medical significance only a small portion, less than 1%, of transmembrane (TM) proteins have a well determined 3-dimensional structure. A number of algorithms designed to identify TM helices in the primary amino acid sequence have been developed. Here, a novel computational approach for the problem of secondary structure prediction in TM proteins with a single TM segment is proposed, by applying the continuous wavelet transform (CWT) to the hydrophobic sequence of the protein. The normalised energy of the resulting signal across all scales of analysis is scanned for continuity allowing for the accurate detection of a single TM segment within the protein sequence. The prediction of both starting and ending point of the TM within the sequence is performed by means of edge detection of the analysed sequence across all scales of wavelet analysis. The scheme was applied to proteins of known structure with a single TM segment extracted by public available databases followed by a comparison with existing methods. Validation results compare favourably with existing methods constituting a promising bed-set for extending the proposed method to the case of multiple transmembrane helix prediction.