Multiscale models are important tools to elucidate how small changes in local subunit conformations may propagate to affect the properties of macromolecular complexes. We review recent advances in coarse-graining methods for poly-protein assemblies, systems that are composed of many copies of relatively few components, with a particular focus on viral capsids and cytoskeletal filaments. These methods aregrouped into two broad categories — mapping methods, which use information from one scale of representation to parameterize a lower resolution model, and bridging methods, which repeatedly connect different scales during simulation —and we provide examples of both classes at different levels of complexity. Collectively, these models illustrate the numerous approaches to information transfer between scales and demonstrate that the complexity required of the model depends in general on the nature of the information sought.