Antiviral discovery using sparse datasets by integrating experiments, molecular simulations, and machine learning

<h2>Summary</h2> Computational methods have proven effective in identifying virucidal agents, contributing to the discovery of novel molecules. In this study, we successfully developed a machine learning (ML) model, trained on a small and imbalanced dataset, to predict inhibitors of human enterovirus 71 (EV71), a significant pathogen that causes severe illness in children and immunocompromised adults. Despite the compact size of the dataset—36 compounds, 5 of which were active against EV71—our ML framework demonstrated significant predictive accuracy. Experimental validation confirmed that five out of eight compounds predicted from the Chinese cosmetic material list exhibited virucidal activity. Molecular dynamics simulations further validated the inhibitory effects of the most active compounds. These findings underscore the potential of our AI-driven approach to accelerate the identification of active molecules against viral pathogens, even with limited data.