A coarse graining method for the dimension reduction of the state space of biomolecules

Abstract

The simulation of biomolecules requires a vast amount of computer power. One reason can be found in the high dimensionality of the state space. Coarse graining methods attempt to improve the computational performance by reducing the representation of a molecule’s dynamics without losing relevant details. Here, we show two methods coarsening the full model by still retaining the details which are necessary for the comprehension of the protein’s conformational dynamics. We review the first method, which clusters motions of the particles according to a certain criterion. This approach is a coarse graining strategy in time based on Markov State Models. The second method, the Hierarchical Relevant Descriptor Detector, is a novel technique for coarse graining in space revealing a hierarchy in the descriptors of a protein. Thus, it allows us to describe the relevant motions of a molecule by employing only a minor number of descriptors. The performance of this method is shown in two examples.