Identification of potential MenT3 inhibitors for Mycobacterium tuberculosis using the generative artificial intelligence and SilicoXplore platform.

Tuberculosis (TB) is the leading cause of death from a single infectious agent, with rising multidrug resistance undermining current treatments. The mycobacterial toxin MenT3 inhibits protein synthesis by covalently attaching CMP to the 3′-CCA end of tRNA, promoting persistence under stress and serving as a promising new target. In this study, a combined machine learning (ML) and physics-based virtual screening pipeline was used to identify MenT3 inhibitors. Approximately 100,000 de novo compounds were generated using REINVENT4, then sequentially filtered by ADMET-AI and PharmacoNet, retaining 11,625 and 1724 molecules, respectively. Triple-replicate docking identified 1481 hits with higher affinity than the co-crystallized ligand. Combining similarity searching with the pyrimidine-ring requirement shortlisted 14 candidates. Initial 20 ns molecular dynamics (MD) simulations and extended MD with MM-GBSA calculations on the top five compounds confirmed superior, stable MenT3 binding. The density functional theory (DFT) study showed that the top molecules exhibit favorable properties, including increased reactivity and stability, optimal charge distribution, and better thermodynamic stability than the reference compound, CTP. These five molecules might be promising for MenT3 inhibition and deserve experimental validation as next-generation anti-TB agents.