A New Method for Identifying Essential Proteins by Measuring Co-Expression and Functional Similarity

Identification of essential proteins is a fundamental task for understanding cellular life. With the increasing availability of high-throughput data, which enable the identification of essential proteins by computational methods from the network level. Various computational methods have been proposed based on topological properties of protein-protein interaction (PPI) network or combining additional biological information. However, the prediction precision is still unsatisfied especially when predicting a small amount of essential proteins. In this paper, we propose a novel method for predicting essential proteins by integrating Gene expression profiles and Gene Ontology (GO) annotation data, called GEG. To demonstrate the performance of GEG method, we evaluated GEG on two PPI networks of Saccharomyces cerevisiae. Simulation results showed that GEG achieved better result performance than the five other state-of-the-art methods. We also demonstrate that the GEG is robust against perturbation in terms of precision-recall and receiver operating characteristic measures. The results indicate that appropriate integrating topological properties with additional biological information will be a great help for identification of essential proteins. The new proposed method GEG is effective and useful for predicting essential proteins in PPI networks.