Prediction Of Blood-secretory Human Proteins Using Support Vector Machines And Increment Of Diversity

The protein of syntheticed in a cell that secreted into the extra-cellular space by various secretion pathways, this protein is called secretory protein. The phenomenon of protein’s secretory is one of the common biological process, it is important that organism guarantee the basic vital activities. Blood-secretory human protein is the one class of secretory protein, which associated with different developmental stages of cancers, suggesting that they could possibly be used as markers of both cancer typing and staging, and developed new drugs. Thus, the effective methods for predicting the blood-secretory human protein become increasingly important in the field of bioinformatics research. Identification of blood-secretory human protein is a very important problem, but it is rather challenging due to the composition complexity and the large dynamic range of proteins in human sera.In this thesis, we investigated the prediction of blood-secretory human protein by various methods. The main contributions are summarized as follows: 1. With the continuous renewal of protein database, we constructed a new blood-secretory human protein database enlarged the number of blood-secretory human protein sequence.2. Three kinds of algorithms are used to predict the blood-secretory human protein:support vector machine (SVM), increment of diversity (ID) and the increment of diversity combined with the support vector machine (IDSVM). The results show that they are all effective methods and obtained the high predictive success rates.3. We proposed a new feature extraction method:average chemical shift information, and combined the average chemical shift with other information of protein sequence. These information parameters are combined with ID, SVM and IDSVM to predict the blood-secretory human protein. The results show that the parameter of considering average chemical shift can obtain higher predictive success rate, the predictive success rate is15%higher than that only considering amino acid composition feature in Jackknife test.