The Research And Establishment Of Algorithm Model Of The Protein Spatial Structure Similarity

Protein is the material basis of organisms and embodiment of biological function. Protein almost involves in all of the important component parts of the organisms. Understanding the three-dimensional structures of proteins can effectively reveal the function of a protein. Protein ’ s structure and function are unified and its biological function is largely dependent on its spatial structure.Proteins with Similar structures often have Similar functions. Study on protein structure and function not only helps people understand the acting mechanism of protein, but has great significance in studying the law of life development.The paper introduced the method of establishing the model of comparing protein spatial structure Similarity. Firstly, coordinates center of the comparison proteins was found to establish the spherical polar coordinate system. According to the radius of the sphere, the azimuth and elevation, the protein region was partitioned, forming a space partition like spherical shell fragments. For comparison,we divided the sphere radius, azimuth and elevation into 2 and 3 parts equally,namely, the protein was divided into 8 and 27 blocks for calculation. With some bioinformatics experts’ suggestions, this study calculated 12 characteristic parameters were selected for analysis, and the 12 characteristic parameters are the density Similarity of each region, P and S position Similarity, atom number Similarity, number and type of amino acid Similarity, number of C, N, O, H, P and S atoms Similarity. For comparison, we adopted SPSS to establish the full regression model of Sim and Sim1~Sim12 in halving or trisection. As a comparison, we also established a BP neural network model to calculate the Similarity of protein.The innovation points of the study can be listed as follows:1. The paper introduced the method of establishing the model of comparing protein spatial structure Similarity. We divided the sphere radius, azimuth and elevation into 2 and 3 parts equally, namely, the protein was divided into 8 and 27 blocks for calculation.2. 12 characteristic parameters were selected, which effectively combined with the protein spatial structure and biological characteristics.3. In this study, the linear and nonlinear regression models of BP neural network were established to compare.Additionally, establishment of the mathematical model can be applied to new protein Similarity comparison algorithm and make up the situation where the BLAST is unable to get ideal results of sequence Similarity comparison sometimes.