Analysis And Prediction Of Protein-protein Interactions And Design Of Management Tool

Currently, research on life science has entered a post-genomics era, meanwhile the major part of life science has been focused on structural genomics and proteomics. As we known, most proteins perform function by interacting with other proteins. As a result, the research on Protein-Protein Interactions(PPI) is becoming more and more important.With the rapid development and application of the technology of the high-throughput biological experiments, a large number of protein-protein interaction experimental data has been produced. However, the results of this biological method are blocked by the high rate of false positive data and negative data. So, the effective calculation method is used to predict the protein-protein interaction. This paper aims to use the theory of machine learning and pattern recognition to predict protein-protein interaction based on protein sequence information.The major content of this paper is separated two main parts, the algorithm of the protein-protein interactions and the software-tool managing the data of protein-protein interactions. In this paper, the innovation and content can be seen as the two following aspects.1,The algorithm of the protein-protein interactions in this paper is based on the information of the proteins'sequence. Firstly of all, we need to deal with the sample data in that the form we needed is special , which can fit our programs well. Secondly, we extract the feature on the amino acid frequency, location, physical and chemical properties, biochemical characteristics of similarity from the protein sequence. Thirdly, the Support Vector Machine(SVM) , which supports small sample classification and prediction, is used as the classifier because the SVM is based on structural risk minimization of statistical learning theory. The models of the SVM based on the features make up the ensemble classifier. We use the ensemble classifier to predict the final results. In this paper, the ensemble classifier algorithm is studied in detail, and we calculate and deduce the theory constraints of the ensemble classifier. On the three datasets, human, yeast and drosophila, the algorithm is verified. Finally the results of the prediction is partly higher than the literature predicted results.2,The second part of the paper is focused on the design of the PPI data management tool. We design the tool as the result of the complex of the original protein sequence. So wide variety of the data sources may cause the difficult of the query, insertion and management. However, the most existing network database management systems do not provide the information of the protein sequence and the information of the protein-protein interaction at the same time. You need to search two or more database systems to obtain the data of PPI and sequence at the same time. Obviously, it is very inconvenient when you do your research on PPI. In view of this situation, we design the data management tool to meet this short, and our tool can help the user to manage their PPI data by the convenient insertion, delete and selection. The most important thing is that we combine the information of the sequence and the information of the PPI in the tool. This structure of the tool is B/S, and the prospect of this tool is based on ASP.net Web development platform and HTML scripts, the background using C#.net program to complete the three message processing functions. The communication between the browser and server is based on the API functions. We use the SQL server to store the data in the tool. This tool in the paper is an effective attempt to the protein-protein data management, and the function of this tool can totally meet the real need in this paper.