Backmapping of Highly Coarse-Grained Protein Models and Applications to Nanomedicine Designs

Multiscale molecular modeling is a powerful tool for computationally studying the structures and dynamics of complex assemblies in chemistry. However, reconstructing all-atom (AA) structures from highly coarse-grained (HCG) models remains a substantial challenge, limiting our ability to obtain essential atomic details at mesoscopic temporal and spatial scales. Within the hierarchical modeling theory and framework, we introduce a progressive backmapping strategy that reconstructs AA models gradually across neighboring resolutions, such as from a 3-residue-per-site HCG model to a 1-residue-per-site model, then to an AA model utilizing a neural-network-based approach called ProNet Backmapping. This progressive strategy achieves accurate reconstructions across various proteins and effectively reconstructs flexible linkers within multidomain architectures. It supports hierarchical reconstruction of complex protein assemblies, including multiple virus-like particles spanning tens of nanometers and containing hundreds of subunits. Notably, we demonstrate the ability to hierarchically backmap entire viral assemblies from HCG to full AA resolution across at least three different levels. This framework enables backmapping from a generic HCG model to various chemical variations, including mutations and modifications, and can be reliably used in future nanomedicine design. Overall, our methodology provides a highly scalable and rigorous framework for incorporating atomistic detail into mesoscale simulations of complex systems across many fields of chemistry.