Integrated framework for targeting dynamic penicillin-binding protein 2a via ensemble structural biology and deep generative modeling.

Methicillin-resistant Staphylococcus aureus (MRSA) remains a major cause of morbidity and mortality worldwide, largely due to penicillin-binding protein 2a (PBP2A), a low-affinity transpeptidase encoded by mecA. To address the challenge of this highly flexible target, we developed an integrated discovery framework combining conformational ensemble analysis with deep generative modeling. Extensive molecular dynamics simulations were used to map the dynamic landscape of PBP2A and identify cryptic binding pockets. A target-conditioned generative workflow then explored these states to produce novel candidate scaffolds, followed by multi-objective optimization and structural prioritization. Among the generated molecules, Compound 1 showed a conserved binding mode, sustained stability over 1000 ns MD simulations, and more favorable MMGBSA binding energies than native ligands. Whole-cell assays showed dose-dependent growth inhibition against S. aureus ATCC 29213 and MRSA ATCC 43300, with stronger activity after 24 h exposure. Additional propidium iodide staining revealed a concentration-dependent increase in PI uptake in both strains, consistent with compound-induced envelope perturbation. These findings support Compound 1 as an antibacterial lead emerging from a PBP2A-guided discovery strategy, while direct target engagement remains to be established.