Generative Adversarial Network-Based Data Augmentation Method for Anti-coronavirus Peptides Prediction.

The Coronavirus Disease 2019 (COVID-19) pandemic has induced a serious public health threat worldwide. Anti-coronavirus peptides (ACVPs) could act as potential peptide drugs for COVID-19. Computational methods for ACVP identification can improve the development of COVID-19 therapeutic. In this work, we propose PredACVP, which is an ACVPs prediction model using generative adversarial network (GAN)-based data augmentation method and stacked ensemble learning. The GAN-based data augmentation method is utilized to overcome the few-shot learning problem and improve the prediction performance. With the advantage of converting a high-dimensional vector into a low-dimensional vector, the stacked ensemble learning could fuse multi-view information (amino acid composition, dipeptide composition, composition of k-spaced amino acid group pairs, and physicochemical properties) without overfitting. The PredACVP model, which achieves AUC of 0.990 on test datasets, outperforms the state-of-the-art tools for ACVPs identification. PredACVP can improve the prediction performance, and accelerate the development of peptide drugs for COVID-19. GAN used in this work not only can be applied on the data augmentation and peptide for ACVP sequences, but also can be utilized for the other therapeutic peptides.