A new protein classification method using dynamic classifier

Since functions of protein may come from its structure, the method of measuring structural similarity between two proteins can infer their functional closeness [1]. Fast structural comparisons and retrieval methods are necessary to deal with the increasing number of protein structural data via high-throughput structural genomics research. Many structural comparison methods have been proposed [2][3]. Among them, distance matrices, approximation of structure, and vector representation are the most commonly used. The distance matrix, also called distance plot or distance map, contains all the pair-wise distances between alpha-carbon atoms, i.e. Cα atoms of each residue. A matrix is a two-dimensional (2D) representation of a three-dimensional (3D) structure. This representation has critical weak points in terms of computational complexity and sensitivity to errors in the global optimization of alignment. In recent research work, this representation was improved by taking into consideration average conformations, the average coordinate of a small number of contiguous residues, instead of alpha-carbon atoms. The improved representation is an approximation of the protein structure for a quick comparison. It overcomes the weak points on sensitivity to errors and computational complexity. In this paper, we designed a protein classification model using the dynamic classifier and described, in detail, all processes in the model. We, then, designed and implemented a protein structure classification system based on the dynamic classifier.