AOEN: A Computational Method for Identifying Essential Proteins by Integrating Protein Abundance, Orthology and Weighted Network Topology

Essential proteins are critical for cell survival and normal physiological activities, whose identification contributes to the exploration of cellular mechanisms and drug-target screening. Experimental biological approaches for detecting essential proteins are accurate but suffer from high cost, long experimental cycles and unavoidable noise. Therefore, computational prediction methods based on protein-protein interaction (PPI) networks have become an important complementary strategy. Nevertheless, most existing topology-based methods rely heavily on static PPI data, and fail to sufficiently combine evolutionary conservation and protein abundance information, limiting prediction performance. In this work, we propose AOEN, a novel computational method for essential protein identification. First, a weighted PPI network is constructed by fusing gene co-expression similarity, GO functional similarity and subcellular localization consistency. Next, individual biological attributes are quantified using protein abundance and orthologous conservation information. Network global and local topological features including eigenvector centrality are extracted from the weighted network. The final score is obtained by combining biological attributes with topological characteristics. We validate AOEN on two Saccharomyces cerevisiae PPI datasets (DIP and Krogan) and compare it with twelve competing approaches. The jackknife curve, precision-recall curve, ROC and multiple quantitative metrics demonstrate that AOEN achieves superior prediction accuracy and robustness. The proposed method provides an alternative tool for identifying essential proteins, and it can offer references for related biological research.