Essential Protein Identification Algorithm Based on Multi-order Heterogeneous Graph and Hybrid Walk

The absence of essential proteins can lead to cessation of cell replication or death, so accurate identification of essential proteins is crucial for understanding cell mechanisms and developing targeted drugs. However, current computational methods mostly focus on binary interaction networks, and do not fully consider the high-order topological characteristics of proteins in protein interaction networks (PIN). To solve this problem, we propose a multi-layer heterogeneous network combined with random walks to identify essential proteins. First, high-order structures are extracted from PINs and protein complexes and combined with PINs and GO annotations to form a multi-order heterogeneous network. Secondly, a hybrid walk strategy is designed for multi-order heterogeneous networks. Finally, the importance of proteins is ranked according to the state distribution at the time of walk convergence. Experiments have shown that this method has better performance than traditional centrality methods.