Frontiers in Coarse-Graining: From Machine Learned and Reactive Molecular Dynamics to HIV-1 Capsid Uncoating and Nuclear Entry

Molecular Dynamics (MD) is a powerful computational method for simulating complex chemical, materials, and biophysical systems. In MD, Newton’s equations of motion are used to describe the dynamic interactions and energetics between atoms, or "particles", and it is the collective motion of an ensemble of particles that enables the simulation of real-world systems such as cell membranes and viral capsids. However, despite its widespread use and the large spatiotemporal scales accessible to classical MD, it remains limited by its fine-grained representation of matter, which makes simulations of systems containing millions of atoms (or more) beyond the microsecond timescale computationally prohibitive. Furthermore, MD based on classical empirical potentials lacks an explicit description of electronic structure and therefore cannot accurately model quantum effects or chemical reactions.

This thesis aims to bridge these limitations by combining coarse-graining (CG) with multiscale reactive molecular dynamics, thereby enabling the efficient simulation of chemical reactions in large-scale systems. In addition, methods that integrate CG with modern machine learned potentials are developed to further improve computational efficiency while retaining high accuracy. Beyond methodological developments, this thesis also advances the modeling and mechanistic understanding of key stages of the HIV-1 life cycle. In particular, computational frameworks are developed to accurately model capsid uncoating as a direct and dynamic consequence of reverse transcription within the viral capsid. Furthermore, extensive all-atom MD simulations provide mechanistic and structural insights into the roles of cellular host factors in guiding viral capsid translocation through the nuclear pore complex during infection. Overall, this thesis presents a comprehensive body of work encompassing methodological advances, state-of-the-art CG model development, and their application to complex biophysical systems, thereby contributing to pushing the frontiers of CG MD.