Prediction Of Protein-protein Interaction Sites Based On AdaBoost Method

With the human genome project and other species genomic sequence determination of s cheme successfully completed, human study of biology from genome era into the post geno mic era. The prediction of protein-protein interaction sites can lead to better understanding of the molecular recognition process of proteins, thus providing valuable information for drug design and application to disease treatment.Therefore, the research on the prediction of prote in-protein interaction sites is of great practical significance for human beings.In this paper, we predict the interaction of protein binding sites through researching on the hotspot residues of protein-protein interaction, using the characteristics of amino acid physicochemical properties, protein structure information and so on to constitute the related feature set to predict hotspot residues. In this paper, we select F-score evaluation method in feature selection. Ada Boost method and SVM are used to construct the forecasting model, to predict the hotspot residues and the non-hotspot residues.Finally, we compared the result on the testing dataset with other models, the prediction model we use in the evaluation index of the Recall value can be reached 53.4%, F1 value reaches up to 51.2%, superior to other prediction model of evaluation index, verify our method is feasible.