Protein Sequence Matching Using Parametric Spectral Estimate Scheme

Putative protein sequences decoded from the messenger ribonucleic acid (mRNA) sequences are composed of twenty amino acids with different physical-chemical properties, such as hydrophobicity and hydrophilicity (uncharged, positively charged or negatively charged amino acids). In this paper, the power spectral estimate (PSE) technique for random processes is applied to the protein sequence matching framework. First, the twenty kinds of amino acids are classified based on their hydrophobicity and hydrophilicity. Then each amino acid in the protein sequence is mapped to a corresponding complex value. Consider the various Hidden Markov chain orders in the complex valued sequences. The PSE method can explore the implicit statistical relations among protein sequences. The mean squared error between the power spectra of two sequences is determined and then used to measure their similarity. The experimental results verify that the proposed PSE method provides the consistent similarity measurement with the wellknown ClustalW and BLASTp schemes. Moreover, the proposed PSE can show better similarity relevance than ClustalW and BLASTp schemes. Keywords—protein sequence; amino acids; digital signal processing; parametric spectral estimate; hydrophilicity; hydrophobicity; Markov chain