AI-enabled discovery of small molecule β-lactamase inhibitors to overcome enzyme diversity and resistance evolution.

Artificial intelligence (AI) is reshaping modern drug discovery by enabling data-driven exploration of chemical space, predictive modeling of molecular interactions, and multi-objective optimization of therapeutic candidates. These capabilities are particularly important in addressing antimicrobial resistance, where conventional discovery strategies have struggled to keep pace with rapidly evolving bacterial enzymes. Among the most urgent threats are carbapenemases, which undermine the clinical efficacy of carbapenems long considered last resort antibiotics for multidrug-resistant infections. Carbapenemases, including metallo-β-lactamases (MBLs) and serine β-lactamases (SBLs), display extensive mechanistic and structural diversity spanning Ambler classes A-D, increasingly accompanied by dual carbapenemase producers. This heterogeneity fundamentally constrains traditional inhibitor discovery and limits the development of resistance-resilient therapeutics. This review provides a forward-looking perspective on the integration of AI, generative molecular design, and structure-informed computational modeling for the discovery and optimization of small molecule β-lactamase inhibitors (BLIs). We examine established and emerging chemical spaces including diazabicyclooctanes, cyclic boronates, and non-β-lactam scaffolds and discuss how AI-driven frameworks are transforming their design. Emphasis is placed on neural networks, reinforcement learning, and synthesis-aware multi-objective optimization for navigating enzyme diversity and anticipating resistance trajectories. We further highlight the role of structure-based modeling, molecular dynamics, and AI-enabled ADMET prediction in addressing active site plasticity, metal coordination, and cross-class binding constraints. Collectively, AI is positioned as an integrative, resistance aware paradigm for the rational development of next-generation BLIs.