Analyzing Similarity Of Protein Sequences With Discrete Wavelet Transform

To find the functional similarity of proteins with low identity is a hinge of protein sequence analysis. Wavelet transform shows the advantage in finding the homology proteins with low similarity, especially whose identity is lower than 25 %. In order to standardize this method, five factors of analyzing the similarity of protein sequences with discrete wavelet transform (DWT) were discussed, including the substitution model, the wavelets, the decomposition levels, the comparing method, and the evaluation of similarity. (i) In this paper, three substitution models were compared. (ii) The better wavelets and the decomposition level fitting for analyzing protein sequences were studied. (iii) The length of sequences affects the result more than the degree of identity when using DWT to analyze the similarity of two protein sequences. To minimize the effects of the sequence length, the segmentation comparing method was designed. The results indicate that the segmentation strategy has greatly improved the performance of analyzing the long protein sequences. (iv) A new evaluation 'S' wasdesigned to measure the similarity of sequences, which was successfully used as a supplement in protein sequence analysis.