Wavelet Transform Based Protein Decoy Discrimination

A decoy discrimination method using wavelet analysis is presented in this paper. Through discrete wavelet decomposition and reconstruction, low frequencies of physicochemical properties from amino acid sequences and solvent accessibility area of residues from the protein three dimensional conformations are extracted and compared to find the profile similarity between the protein sequence and its three dimensional conformation. With the assumption that the protein native state conformation should have more profile similarity to its sequence than any other decoys, a scoring function evaluating the similarity is developed. This method was trained and tested by the first_decoys set in rosetta decoy set. The test results show that 73% protein native structure has the top 5 largest score. Considering the simplicity of our method, wavelet analysis is a promising way to deal with the decoy discrimination problem.