Integrative bioinformatics, machine learning, and molecular docking identify HigBA toxin-antitoxin systems as key...

Acinetobacter baumannii is a high-priority pathogen due to its extensive antimicrobial resistance and persistence in clinical environments. Quorum sensing (QS) and toxin-antitoxin (TA) systems regulate virulence and stress tolerance, yet their interconnection remains unclear. We analyzed transcriptomic data (GSE87009) from A. baumannii treated with 3-oxo-C12-HSL using WGCNA, differential expression, and machine learning (LASSO and Random Forest). To identify potential inhibitors of the HigBA TA system, we performed molecular docking of 202 chemically diverse compounds - comprising approved drugs, natural products, and synthetic molecules - against the four HigBA proteins using AutoDock Vina, followed by Protein-Ligand Interaction Profiler (PLIP)-based interaction profiling to validate binding modes. WGCNA identified Module 1 (10 genes) as the primary QS-responsive module (r = -0.996, p = 0.0041), containing two complete HigBA systems (chromosomal and plasmid-borne), proteases, transposases, and a catalase. LASSO and Random Forest converged on the four TA genes as robust QS predictors. Docking identified Ligand 17 as the best binder for HigB2 (-7.65 kcal/mol), followed by Ligand 147 for HigA2 (-6.61 kcal/mol), Ligand 198 for HigB1 (-6.79 kcal/mol), and Ligand 178 for HigA1 (-4.51 kcal/mol). PLIP validation confirmed that all ligands occupy the ATP-binding cleft; Deslanoside (147) forms 23 hydrogen bonds, Meclizine (198) anchors via CYS83, while Promazine (178) and Estradiol (17) rely on hydrophobic contacts. These findings provide a mechanistic link between QS and TA systems, thus providing new therapeutic targets for tackling virulence and persistence.