Computational strategies for antimicrobial discovery: From machine learning to multiscale simulation

Antimicrobial resistance (AMR) has become a major threat to global public health, and there is an urgent need to develop innovative antimicrobial research and development strategies. Traditional experimental screening methods struggle to meet the current needs due to their high cost and long development time. The rise of computational biology brings new hope to antimicrobial discovery, with the “computation-first” strategy being potentially able to effectively accelerate antimicrobial research and development. This article systematically reviews key technologies in the field of computational biology, such as machine learning (ML) technology, molecular dynamics (MD) simulation technology, and hybrid artificial intelligence-molecular dynamics platforms, and thoroughly discusses the related challenges in clinical translation, such as those associated with discrepancies in bioavailability between virtual screening predictions and experimental validation results as well as in toxicity predictions between in vitro and in vivo experiments. Finally, this paper proposes the integration of multidisciplinary technologies into a theoretical framework and details policy recommendations for dealing with AMR. Specifically, we propose a “computation–experiment–clinical translation” closed-loop framework that integrates ML-driven design, MD-based mechanistic validation, and feedback based on real-world clinical data, thus addressing the fragmentation of current research pipelines.