Research Projects

Flaviviruses, including dengue (DENV), Zika (ZIKV), West Nile (WNV), and Japanese encephalitis (JEV), pose a significant global health challenge due to frequent ­outbreaks, high morbidity in severe disease, and the lack of effective antiviral therapies. Existing vaccines, the diversity of viral serotypes and the severity of infections underscore the critical need for broad-spectrum antivirals targeting conserved mechanisms. One particularly promising target is the NS3 helicase, a highly conserved enzyme essential for viral RNA replication through its roles in unwinding double-stranded RNA, catalyzing ATP hydrolysis, and mediating interactions with other viral proteins. Structural studies have revealed that NS3 helicase possesses conserved active sites and allosteric pockets, including the ATP binding cleft and RNA recognition interface, indispensable for its enzymatic function. Despite its pivotal role in the viral life cycle, the helicase remains underexploited as a drug target, mainly due to challenges arising from its dynamic conformations and the difficulty in identifying high-affinity inhibitors. To address this gap, our project proposes a Machine Learning (ML)-assisted strategy to identify and optimize broad-spectrum antiviral agents targeting shared structural and functional motifs within the NS3 helicase domain, thereby enabling the rational design of inhibitors effective against multiple flavivirus species. This approach leverages high-resolution Protein Data Bank (PDB) structure to conduct detailed structural analyses to identify conserved binding pockets, flexible regions, and druggable allosteric sites. Subsequently, we will employ state-of-the-art physics-based virtual screening protocols such as RosettaGenFF-VS/Rosetta GALigandDock to efficiently explore vast chemical spaces, while incorporating target flexibility by modeling sidechain and limited backbone movements to capture induced-fit effects characteristic of dynamic NS3 helicases. Screening will be performed on billion compound libraries using OpenVS, an AI-accelerated, scalable virtual screening platform that integrates ML-driven active learning to enhance compound selection. In parallel, tailored ML-models will be developed and trained on physicochemical, topological, and dynamic descriptors derived from known NS3 helicase inhibitors to predict binding affinity and prioritize high-potential candidates for further refinement. Top hits will be subjected to MD simulations and free energy calculations to evaluate binding interactions and inhibitory potency. The pipelines integrates bioinformatics, advanced docking methods, and ML-assisted screening to accelerate the discovery potent pan-flaviviral antiviral. Successful implementation of this approach will deliver a new class of broad-spectrum antiviral and provide a reproducible model for applying ML-assisted atomistic simulations to other dynamic viral targets, thereby transforming the landscape of antiviral drug discovery.