Breaking resistance with machine and deep learning: A computational intelligence hunt for AmvR (TetR) inhibitors in...

Multidrug-resistant Acinetobacter baumannii infections have driven the development of innovative therapeutic approaches to address global challenges. The present study reported that the protein belonging to the TetR family regulator AmvR from A. baumannii is bound to spermidine. Here, quantitative structure-activity relationship (QSAR) models developed from a ChEMBL dataset were used to perform an initial screening of a library of 10,860 natural chemical compounds, along with their reported minimum inhibitory concentration (MIC) scores. Machine learning models, including RF, SVM, KNN, and XGBoost, were implemented for compound classification. An RF model was selected that achieved a 96 % accuracy score on the test set. A deep learning model, CNN, was built, and a structure-based virtual screening approach was applied against A. baumannii to assess how similar these molecules are to drug likeness. In silico pharmacokinetic analysis was performed for the top candidates, including binding energy evaluation of protein-ligand docking with scores of -10.6 kcal/mol, -10.4 kcal/mol, and a control with -6 kcal/mol. Absorption, distribution, metabolism, and excretion (ADME) properties for CHEMBL-560977, CHEMBL-551116, and the control complex were also noted. Additionally, the binding mechanism was elucidated through molecular dynamics simulations over 500 ns, with binding free energy analysis (MMGBSA/PBSA) showing results of -120.55 kcal/mol, -94.99 kcal/mol, and -90.65 kcal/mol. Dynamic cross-correlation matrix (DCCM), radial distribution function (RDF), and hydrogen bonding (H-bonds) analyses were conducted, along with principal component analysis (PCA) and free energy landscape (FEL) clustering for the selected complexes. Secondary structure analysis and salt bridge assessments indicated that the newly discovered compounds are more promising than the currently available drugs. According to all docking and MD simulation results, the ligand-bound protein exhibits promising and stable behavior. Lastly, the selected compounds are recommended for further experimental studies and could serve as potent inhibitors of A. baumannii infection.