Integrated deep learning and multi-scale modeling for the discovery of pan-genotypic HCV NS5B polymerase inhibitors.

Chronic Hepatitis C Virus (HCV) infection remains a significant global health challenge, necessitating the development of pan-genotypic inhibitors to overcome the limitations of genotype-specific therapies. The NS5B RNA-dependent RNA polymerase, particularly its conserved Palm II region, is a promising target for such broad-spectrum antivirals. In this study, we integrated advanced deep learning and multi-scale computational approaches to design novel pan-genotypic HCV NS5B inhibitors. First, we employed AlphaFold3 to predict the high-resolution structures of NS5B polymerase from genotypes with previously unresolved crystal structures (GT3a, GT3b, GT6a, GT7a, and GT8a), establishing a robust structural database refined by 100 ns molecular dynamics (MD) simulations. Concurrently, the DrugEx algorithm, a multi-objective reinforcement learning-based molecular generator, was used for de novo design, yielding 42,257 novel chemical entities. Subsequent multi-step virtual screening, molecular docking, and 300 ns MD simulations identified four hit compounds (01, 07, 18, 59) that demonstrated promising computational binding profiles and pan-genotypic coverage compared to the positive control BMS-986,139. These compounds exhibited excellent stability in complex with NS5B across ten genotypes, as evidenced by low RMSD fluctuations. MM/GBSA binding free energy calculations and hydrogen bond analysis revealed key interaction mechanisms, including dynamic hydrogen bonding with Tyr448 and strong electrostatic contributions from Arg200, which underpin their broad-spectrum efficacy. Furthermore, ADMET predictions confirmed favorable drug-like properties and synthetic feasibility, supported by retrosynthetic analysis. This work presents a comprehensive "AlphaFold3-DrugEx-Multi-scale Simulation" strategy, enabling the efficient discovery of novel, potent, and pan-genotypic HCV NS5B inhibitors candidates with significant potential for further development.