Using K-Nearest Neighbor And Multiple Improved Methord To Identify Anti- And PRO-Apoptosis Proteins

Protein is an important component part for body’s tissue and organs, and which has intricate biological functions in biological processes; all chemical reactions is catalyzed by protein within the cells; The reason that the body’s life activities is ordered is due to the coordination of protein. The function of protein is also quite different in various cells. There is a special proteins have been a lot of research because of its special function apoptosis proteins. Apoptotic proteins can be subdivided into below two types according to the difference of its function: anti-apoptotic proteins and pro-apoptotic protein.This paper used the dataset of anti-apoptotic proteins and the pro-apoptotic protein that builded in 2013 to evaluate proposed and improved algorithm, and then analyzed the classification results of multiple algorithm. During the process of evaluating algorithm and analyzing result, we extract the following feature information; amino acid composition information, component segment information, physical chemistry information, evolutionary information, the gene ontology information based on bioprocess and molecular function and so on. We proposed a reduction dimension method of fusion Support Vector Machines(SVM) and Shannon information for high-dimensional feature space. As we know SVM algorithm is recognized as relatively stable classifier, so in order to estimated performance proposed and improved algorithms, we adjust SVM to optimal state for comparison; Then the feature space will be input to algorithm used (K-Nearest Neighbor (KNN), Increment of Diversity (ID); fusion of K-Nearest Neighbor and Increment of Diversity (ID-KNN)) and improved algorithm to predict, and evaluate the algorithm based on the prediction results. We can find out that different algorithms obvious difference for different feature information. And during result’s analysis, the Assessment of prediction performance by Jackknife method. We have selected single better feature to fusion and results better than that of single feature.