Machine Learning‐Guided Discovery of Novel Inhibitors of Escherichia coli MurB

ABSTRACT

MurB, a key enzyme involved in bacterial cell wall biosynthesis, represents a promising target for antibacterial drug discovery. In this study, a machine learning assay was employed to identify potential inhibitors targeting E. coli MurB. Five machine learning models were constructed using AutoMolDesigner, of which the RDKit_2D ‐ based Model_2 was selected as the optimal activity ‐ prediction model for virtual screening of an in ‐ house compound library, yielding 238 compounds. These compounds were subsequently subjected to FRED molecular docking, and 16 candidate compounds were selected based on their hydrogen ‐ bond interactions with the key active ‐ site residue ASN A:233. Preliminary enzyme inhibition assays conducted at 50 μM identified 4 compounds (Q9, Q11, Q12, and Q13) with inhibition rates exceeding 50%. Further IC₅₀ determination revealed values of 34.44, 31.45, 25.08, and 25.94 μM, respectively. Molecular dynamics simulations combined with principal component analysis were performed to investigate the dynamic behaviors and binding stability of the MurB ‐ ligand complexes. In addition, a CCK ‐ 8 assay was conducted in HepG2 cells to evaluate the cytotoxicity of the active compounds preliminarily. Collectively, 4 low ‐ cytotoxic micromolar MurB inhibitors were identified in this study, offering promising scaffolds for further mechanistic investigation and structural optimization.