Essential protein prediction based on generative adversarial networks

Essential proteins in cells and organisms play a critical role in maintaining normal life functioning and also provide important support in understanding disease pathogenesis. Although existing machine learning and deep learning methods have made some progress in predicting essential proteins, the use of data augmentation methods to improve model robustness and generalization becomes particularly important due to the critical role of data in training models. However, it remains a challenge to use limited data for data augmentation to improve the accuracy of predicting essential proteins. Therefore, we propose an algorithm for essential protein identification based on generative adversarial networks. First, we input the preprocessed gene expression data into a pre-trained generative adversarial network generator to expand the existing gene expression dataset. Second, features are extracted through the confrontation between the generator and the discriminator in the generative adversarial network and PCA technique is applied to downscale these features to make them more representative. Subsequently, the Node2vec method is applied to capture the rich features in the Protein-Protein Interaction (PPI) networks. Finally, we fuse the extracted gene expression profile features with the features of the PPI network and input them into a deep neural network for classification. Experimental results show that our proposed method has better performance compared to existing methods for predicting essential protein.