Designing Novel Antiviral Peptides Targeting Dengue NS3 Protease via Long Short‐Term Memory Model, Molecular Docking and Molecular Dynamics Simulations

ABSTRACT The dengue virus NS3 protease, which is critical for viral replication, remains an attractive yet challenging therapeutic target. Traditional peptide design approaches based on expert knowledge are time‐consuming and labor‐intensive, whereas artificial intelligence (AI)‐based methods offer a promising alternative to accelerate this process. In this study, we developed a Long Short‐Term Memory (LSTM)‐based deep learning framework with transfer learning to design novel antiviral peptides targeting the dengue virus NS3 protease. A model pretrained on antiviral peptides was fine‐tuned using NS3‐specific sequences, followed by multistage screening. Four peptides (GP81, GP14, GP79, and GP2) exhibited superior predicted binding affinities (−11.6 to −10.9 kcal/mol) compared with the reference inhibitor (−9.4 kcal/mol), forming stable interactions with key catalytic residues (His51, Asp75, and Ser135). Molecular dynamics simulations further confirmed the stability of GP2, which showed the highest number of hydrogen bonds throughout the simulation. MM/GBSA calculations demonstrated that GP2 possessed the most favorable binding free energy (−78.35 kcal/mol) among the four candidates, significantly outperforming the reference inhibitor (−30.43 kcal/mol). Our integrative approach—combining LSTM, docking, and molecular dynamics simulations—provides a robust pipeline for de novo antiviral peptide design, emphasizing AI‐driven strategies to accelerate drug discovery against dengue.