Taking into account the reliability of protein-protein interaction (PPI) network, an uncertain network was constructed and a novel method named essential proteins identification based on uncertain networks (EPU) was developed to identify essential proteins from the uncertain network. The concept of expect density was used to decide whether a subgraph can be represented as an essential module for essential proteins identification. Proteins were ranked through their probabilistic frequency appearing in these predicted modules. Then the ranking scores of these proteins were used to judge whether a protein is essential. Experimental results showed that the EPU algorithm outperforms other essential proteins prediction algorithms and is a special method which is different from others. The results indicated that the theory of uncertain data management is useful for the improvement of robustness in protein-protein interaction networks.
