Machine Learning-Based High-Throughput Screening, Molecular Modeling and Quantum Chemical Analysis to Investigate Mycobacterium tuberculosis MetRS Inhibitors.

Abstract

The increasing drug resistance of Mycobacterium tuberculosis (Mtb) complicates its effective treatment and often leads to severe side effects. This research aims to pinpoint the potential drug candidates targeting Mtb methionyl-tRNA synthetase (MtbMetRS) using in silico techniques. Employing machine learning algorithms, including Random Forest, Extra Trees, and Nu-Support Vector, a voting classifier was built to screen 10 million molecules. A total of 590 molecules were filtered and analyzed for mutagenicity and other toxicities, resulting in 169 candidates for molecular docking. Among these, 1-[4-(1,3-benzodioxol-5-ylmethyl)piperazin-1-yl]-2-phenylsulfanylethanone (L1) and 1-ethyl-6-fluoro-4-oxo-7-(4-pentanoylpiperazin-1-yl)quinoline-3-carboxylic acid (L2) demonstrated strong binding affinities (-12.74 kcal/mol for L1 and -11.83 kcal/mol for L2) and favorable pharmacokinetic properties. MM/PBSA, DFT calculations, and LD₅₀ values supported their stability, reactive nature, and safer toxicity profile, respectively. L1 and L2 are investigated as potential inhibitors of MtbMetRS; however, additional in vitro and in vivo investigations are necessary to confirm these findings.