A method for predicting essential proteins in heterogeneous networks based on discrete quantum walk

Essential proteins are an important material basis for life activities, and their accurate identification is crucial for understanding cellular functions and biological mechanisms. Although there are already various experimental and computational methods for detecting essential proteins, most of them are limited to a single network or local features, making it difficult to integrate heterogeneous network information, and they lack accuracy and robustness in complex networks. Therefore, we propose a heterogeneous network essential protein prediction algorithm based on discrete-time quantum walk (HNDQW). This algorithm first constructs a protein–domain heterogeneous network according to the original protein–protein interaction (PPI) network and the known relationships between proteins and domains. Then, it constructs a quantum weighted transition matrix of the heterogeneous network by integrating network topological features and multi-source biological information. Finally, it constructs a discrete-time quantum walk model and performs quantum walks on the heterogeneous network. The average probability of reaching each vertex is selected to represent the essentiality score of the protein node. By ranking the proteins, the top-ranked proteins are selected as candidate essential proteins. In addition, to evaluate the performance of the algorithm, we compare this method with 13 other different methods on three datasets, DIP, Krogan and Gavin. The experimental results show that this method achieves recognition accuracies of 94.12%, 63.35%, 56.82% and 51.26% among the top 1%, 15%, 20% and 25% of candidate proteins, respectively. It also performs well under multiple evaluation metrics, with prediction accuracy exceeding that of the other 13 methods, and can effectively identify essential proteins.