Abstract:The identification of essential proteins is not only important for understanding organism structure on the molecular level, but also beneficial to drug-target detection and genetic disease prevention. Traditional methods often employ various centrality indices of static protein-protein interaction (PPI) networks and/or gene expression profiles to predict essential proteins. However, the prediction accuracy of most methods still has room to be further improved. In this study, we propose a strategy to increase the prediction accuracy of essential protein identification in three ways. Firstly, RNA-Seq datasets are employed to construct integrated dynamic PPI networks. Using a RNA-Seq dataset is expected to give more accurate predictions than using microarray gene expression profiles. Secondly, a novel integrated dynamic PPI network is constructed by considering both the co-expression pattern and the co-expression level of the RNA-Seq data. Thirdly, a novel two-step strategy is proposed to identify essential proteins from two known centrality indices. Numerical experiments have shown that the proposed strategy can increase the prediction accuracy dramatically, which can be generalized to many existing methods and centrality indices.

Identifying essential proteins based on dynamic protein-protein interaction networks and RNA-Seq datasets