Predicting Essential Proteins Based On Protein Network And Protein Function

The key proteins is the something necessary for survival and reproduction of organism.Its mutation,remove or damage will cause the disability of an organism,or even death.The essential protein identification contributes to understanding the biological processes in system level,discovering disease genes,establishing drug target,diagnosising and treatmenting for diseases and designing drug.So prediction of essential proteins is important for theoretical research and practical application.Moreover,with the development of high-throughput technology in the post-genomic era,a wealth of protein function and protein-protein interaction data have been produced.Consequently,prediction of key proteins also gradually become a new research hot topic.The thesis takes Saccharomyces cerevisiae Proteins as research object,proceeds from network topology,and graph algorithms theory is used to emphatically analyze the classical centricity-mostly measure prediction method of key proteins based on the protein interaction network.Experimental results of six mining algorithm based on protein network topology centricity point its limitation: it can't dig low key proteins.Pointed to this problem,this paper suggest a new predictive factor,protein function.The main original work includes:1)Analyzing theoretically the relationship between protein function and essential proteins and testing this hypothesis by designing and performing experiments.Experimental results prove protein function has strong correlation with key protein.And on this basis,this paper propose a new predictive algorithm,FUN,and analyze the algorithm performance mining the low key proteins.2)On the base of the above work,a theoretical analysis is made on the relationship of these two predictors between protein network and protein function.And experiments are carried out about above analysis.Experimental results find these two predictive factors complement each other well.Under this conclusion,this paper propose a new prediction algorithm,DC_FUN,to integrate these two predictive factors.3)Conducting comparative experiments on yeast protein by using FUN and DC_FUN proposed algorithm in this paper and DC,IC,EC,SC,BC,and CC algorithm to demonstrate the effectiveness and value of these two algorithm.Experimental results in yeast protein show the prediction accuracy of algorithm based on protein function,FUN,is not less than that based on protein network,and it can beenhanced by DC_FUN algorithm integrating these two predictive factors.