Analysis And Prediction Of Protein Binding Sites Based On Structural Data

Protein plays an important part in the life process.Proteins often interact with ligands to achieve their biological functions,such as the transmission of cellular genetic information,cellular metabolism,material transport and signal transduction.Therefore,the study of the interaction between proteins and ligand molecules is of great significance in the discovery of protein function.In this paper,a reliable data set is obtained through analysis and preprocessing,and then the mathematical model of protein structure analysis is constructed by combining 3D geometric calculation method and data mining technology.We have proposed an improved method for describing the surface morphology of proteins,combined with the physical and chemical properties of proteins,the model we constructed acquired significant classification and prediction of DNA binding proteins and RNA binding proteins.In addition,for the complex data of protein and small molecule,the binding site characteristics of protein and small molecule complex were analyzed,and the binding site prediction of small protein molecules was realized.The full text work consists of two parts:(1)DNA/RNA binding protein binding region analysis and classification prediction.Protein-nucleic acid interaction studies are important for understanding life activities.We calculated the molecular volume and surface area around the residues in the binding region for the binding region of the DNA/RNA binding protein,and then calculated the surface morphology around the residues in the binding region.According to the calculation results,the residues were classified into three types: peak,flat and valley.The solvent accessibility and secondary structure characteristics of the protein binding region were further obtained.By comparison,it was found that there were significant differences in the morphological structure,solvent accessibility,and secondary structure of the binding region of the two nucleic acid binding proteins.Based on these characteristics,the SVM classification prediction model is constructed,and the 10-fold cross-validation method is used for classification prediction,and good classification results are obtained.(2)feature mining and classification prediction of protein and small molecule binding sites.The study of protein and small molecule binding sites is of great significance for drug development and design.At present,the traditional experimental method for detecting the binding sites of small molecules of proteins is costly and time consuming.For example,research tools for the development of small molecules of proteins for a particular ligand often have the disadvantages of being inefficient and difficult to promote.In the experiment,we proposed a classification prediction method based on XGBoost model.By analyzing the evolutionary information and physicochemical properties of proteins,high-dimensional sequence features were obtained,and the important characteristics were selected by mean decrease accuracy.The final constructed classification model achieved very significant predictions.