Synthetic antimicrobial peptides generation using recurrent neural networks

Abstract The antimicrobial peptides (AMPs) have taken importance in the development of new antibiotics because of their role as an inhibitor, not only of bacteria but also of viruses, fungi and parasites, among others. Since the discovery of AMPs, thousands have been reported, however, many of them are not suitable for therapeutic applications due to their long amino acid sequences, low antimicrobial potency and high production costs. In this work, we propose to use recurrent neural networks (RNN) with LSTM cells in order to generate more potent and economical peptides. We perform different experiments generating synthetic AMPs between 12 and 20 amino acids. The results show that we can use RNN and improve the generation process compared with the template method.