The method for predicting essential proteins based on protein-protein interaction network is not deep enough to discover biological functions.In order to solve this problem,we utilize the protein complex information and propose an algorithm named RWP based on random walk model combining with the edge clustering coefficients in PPI network to recognize essential proteins.In protein-protein interaction network of Saccharomyces cerevisiae,5 criteria of statistics evaluation criteria such as SN etc were taken to experiment with RWP and five centrality measure methods (DC,etc.) contrastively.The results showed that the number of essential proteins predicted by RWP was more than that predicted by other five centrality measure methods.