A Fivefold Maximum Drug-Likeness Strategy for Prioritizing Antibacterial Candidates Against Escherichia coli

Background/Objectives: Early-stage antibacterial discovery requires balancing activity with broader developability-related properties. This study developed a Fivefold Maximum Drug-Likeness strategy (5F-MDL) as a multidimensional, developability-aware framework for prioritizing antibacterial candidates against Escherichia coli (E. coli). Methods: An ensemble of endpoint-specific deep learning models was used to construct a 33-dimensional predicted property spectrum. Approximately 16 million commercially available molecules were screened by comparing their normalized property profiles with those of clinically approved cephalosporin reference drugs. Fifteen top-ranked candidates were selected for experimental evaluation. Antibacterial activity was assessed by disk diffusion, minimum inhibitory concentration (MIC), and minimum bactericidal concentration (MBC) assays. Molecular docking, molecular dynamics simulations, MM-PBSA analysis, and a Bocillin-FL competition labeling assay were used as target-related supportive analyses. Results: Among the 15 prioritized candidates, three compounds showed measurable antibacterial activity against E. coli ATCC 25922. M2 showed the most favorable in vitro profile among the active candidates, with an MIC of 25.6 µg/mL and an MBC of 51.2 µg/mL, although its potency remained weaker than that of cefuroxime. Docking, molecular dynamics, and MM-PBSA analyses suggested that M2 could maintain a relatively stable noncovalent interaction pattern within the modeled PBP2 pocket, and the Bocillin-FL assay showed that M2 reduced fluorescent probe labeling of PBP2 in vitro. Conclusions: These findings provide preliminary proof-of-concept support for 5F-MDL as a multidimensional prioritization strategy for early-stage antibacterial candidate selection. M2 should be regarded as a preliminary antibacterial hit for further optimization rather than as a validated lead compound. Broader applicability, comparative advantage, and mechanistic relevance require further validation using larger candidate sets, resistant clinical isolates, systematic benchmarking, and direct target-validation experiments.