How artificial intelligence can boost computer-aided drug design in ESKAPE infections

Abstract

Computer-aided drug design (CADD) is a key approach in drug discovery, playing a crucial role in the development of many approved drugs. The integration of artificial intelligence (AI), driven by the increasing availability of data and continuous model refinement, is creating new opportunities for predicting target structures, drug interactions, molecular properties, novel molecules, and drug repurposing [1]. In this study, we demonstrate how combining physics-based and deep-learning approaches in CADD can help redirect existing drugs to combat ESKAPE infections. The World Health Organization has identified antimicrobial resistance as a major global threat, with ESKAPE pathogens, six highly drug-resistant bacteria, being a leading cause of hospital-acquired infections [2]. Among them, the most concerning are the highly resistant Gram-negative bacteria Klebsiella pneumoniae, Pseudomonas aeruginosa, and Acinetobacter baumannii, which pose a serious risk, particularly to critically ill patients. Our research focuses on the LpxM lipid A acyltransferase protein from A. baumannii, which reinforces the bacterium's membrane, increasing its virulence and enhancing infection and transmission rates [3]. Due to its critical role, LpxM is an attractive target for new antimicrobial drugs. To identify potential inhibitors, we performed virtual screening and molecular dynamics simulations using the "World" subset of the ZINC database, which includes drugs approved in major global jurisdictions, along with the crystal structure of LpxM. The top-ranking candidate drugs, based on calculated binding energies, are being selected for experimental evaluation of their effects on enzymatic activity and bacterial fitness.

[1] Pirolli et al. (2023) Sci Rep 13:1494

[2] Mulani et al. (2019) Front Microbiol 10: 539.

[3] Boll et al. (2015) mBio 6(3): e00478-00415