Protein Function Prediction Based On Protein Squence And Protein-Protein Interaction

With the approach of post-genome era, proteomics is becoming an important research domain in the life science. Prediction of protein function is a challenging problem at present in the research of proteomics. Any new breakthrough in this research will be helpful to ex-pounding the change mechanism of organism under physiological or pathological condition directly. What is more, it will be an important assistant to relevant industries such as bio-medical engineering, agbio-tech, etc.On the basis of thorough analysis of existing protein function prediction method, we develop our work on the study of protein function prediction. We propose two new predict-ing models and perform series of experiments. Finally, we get good results. The main con-tributions of this thesis are summarized as follows:(1) Study in the existing protein function prediction techniques. A thorough analysis of protein function prediction from protein sequence,structure and protein-protein interaction is given firstly. Then we take a closer look at the difficulties protein function prediction is fac-ing now and propose the research scheme of this thesis.(2) A combined predicting method from protein-protein interaction networks is de-signed and implemented. The interaction between proteins is one of the most important fea-tures of protein functions. Aimed at this problem, we regard the protein-protein interaction networks as a small world network and use the"Small-World Networks"characteristic to predict protein function. The results indicate that the"Small-World Networks"-based algo-rithm (SWA-BA) can get higher success rate for the number of interacting partners is small. Especially for the number of interacting partners is smaller than 4, the success rate can is 3~4 percent higher than the GO methods. Finally, the combined predicting model associated with the SWN-BA and Go method is proposed. The results of data experiments indicate that the combined model can well be used in the study in protein function prediction from pro-tein-protein interaction network.(3) A predicting model of Encoding Based on Grouped Weight (EBGW) is designed and implemented. The methods on the basis of protein-protein interaction are not available when the proteins are not interacted with other proteins. Therefore, a new predicting model from protein sequence independently is proposed. Based on the idea of coarse-grained description in physics, using EBGW with amino acid physical-chemical characteristics, we turn the pro-tein sequence into a group of vectors to get more information of protein function and infer protein function combined with the Nearest Neighbor Algorithm. The results of data experi-ments indicate that the new model is efficient and reliable to assign function to unknown proteins that are not interacted with others.