Computational structural modeling has the potential of playing an instrumental role in expanding our understanding of biology at the molecular level. If accomplished, it can provide insight into the functionally relevant conformational motions of molecular machines and predict the change in their biological activity upon mutation or interaction with certain pharmaceutical drugs. However, success is not yet entirely at hand because the biologically relevant collective motion of atoms would spontaneously emerge only at time scales not always accessible by standard molecular simulations (molecular dynamics) performed at conventional academic computing facilities. Here, we introduce an algorithmic approach that enables modelers to use customizable hierarchical degrees of freedom to accelerate the exploration of conformational space thereby circumventing major limitations of conventional molecular modeling. This method opens up new research areas for modeling nanosized macromolecules.
Back to Machine Learning for Many-Particle Systems