Conformational landscape of the HBV capsid CAM binding site revealed by all-atom MD simulations and unsupervised learning

The Hepatitis B virus (HBV) remains a major global health concern, infecting more than 296 million people worldwide and causing both acute and chronic liver disease. Although effective vaccines exist, there is no definitive cure for chronic HBV infection, underscoring the urgent need for novel antiviral strategies. The HBV capsid, a protein shell comprising 120 dimeric core protein (Cp) subunits, plays multiple roles throughout the viral life cycle, making it a prime therapeutic target. Small molecules known as capsid assembly modulators (CAMs) bind at inter-dimer interfaces and can misdirect assembly or destabilize the intact capsid, yet the atomistic mechanisms driving CAM-induced disruption remain unclear. Here, we performed all-atom molecular dynamics (MD) simulations of the intact HBV capsid in the presence of different CAMs, generating over 96 million inter-dimer conformations. Beyond traditional coordinate-based analyses, we employ unsupervised learning, robust statistical tests, and computer-vision–inspired approaches to characterize the 3D volumetric landscapes of these interfaces, providing a geometric and statistical view of how CAM-induced perturbations propagate across the capsid. By integrating these analyses with detailed protein–protein and CAM–protein interaction profiles, we uncover how local interface remodeling drives global capsid disruption, offering molecular insights to guide the design of next-generation HBV antivirals.