Artificial intelligence and molecular modelling-assisted drug discovery pipeline for natural inhibitors against DNA gyrase B subunit of Pseudomonas aeruginosa

Abstract

ABSTRACT Antimicrobial resistance (AMR) is a critical global health crisis that demands the discovery of novel antibacterial agents against multidrug-resistant pathogens such as Pseudomonas aeruginosa. In this study, an integrated artificial intelligence (AI)- and molecular modelling-assisted drug discovery pipeline was developed to identify potent DNA gyrase inhibitors from a natural product chemical library. A supervised XGBoost machine learning model was trained using experimentally validated DNA gyrase inhibitors retrieved from the ChEMBL database and achieved a prediction accuracy of 75% with a weighted F1-score of 0.74. The optimised model was subsequently used to screen 95,625 natural product-like compounds from the NPASS database, identifying 11,978 predicted strong actives. After Lipinski-based drug-likeness filtering, 6784 compounds were retained, from which the top 500 were subjected to molecular docking. The top five candidates, including Baicalein, Ibogaine, Meliternatin, Toxicarol, and Urgineanin D, showing binding affinities ranging from −10.4 to −9.8 kcal/mol, were shortlisted for 100-nanosecond molecular dynamics simulations using GROMACS. Trajectory analysis including RMSD, RMSF, Rg, SASA, hydrogen bonding, PCA, and free energy landscape analysis confirmed stable protein–ligand interactions and favourable conformational stability, particularly for Baicalein and Toxicarol complexes. This approach proposes an innovative and effective strategy to find lead compounds against potential receptors.