Computational screening of antimicrobial peptide analogs targeting AdeB efflux transporter protein in Acinetobacter baumannii associated with multidrug resistance.

Overexpression of the AdeABC efflux pump is crucial in conferring multidrug resistance (MDR) in Acinetobacter baumannii (A. baumannii). Its primary subunit, AdeB, recognizes and pumps out aminoglycosides, carbapenems, and several other classes of antibiotics from the cell, making it an attractive target for drug development. There has been substantial interest in utilizing antimicrobial peptides (AMPs) as potential alternatives to traditional antibiotics. Here, we focused on computer-aided drug design methods to identify effective AMPs against the target AdeB protein. First, we used the Joker algorithm to optimize 45 experimentally validated AMPs that can inhibit the growth of A. baumannii. This created a library of 629 peptide analogs, which were evaluated using a supervised machine learning (ML) pipeline based on 117 descriptors. The CatBoost classifier demonstrated the best performance among the tested model, with an accuracy of 96.5%, precision of 97%, recall of 96.2% and specificity of 96%. Furthermore, 126 predicted AMP were screened and filtered based on their physicochemical and safety criteria. Molecular docking analysis of screened peptides with AdeB transporter identified strong binding affinities for eleven peptide analogs. Post-docking, 200 ns molecular dynamics (MD) simulations (including an independent replicate run) supported the structural stability of three peptide analogs (BMAP27-R9, C20DK-R3, and SAAP148-R1). Binding free energy calculation using MM/PBSA indicated improved score of -61.15 kcal/mol for BMAP27-R9, -28.15 kcal/mol for C20DK-R3 and  -41.80 kcal/mol for SAAP148-R1. These findings provided a basis for future studies on efflux-targeting antimicrobial strategies against MDR A. baumannii.