Kinetically-aware Conformational Distances in Molecular Dynamics

In this paper, we present a novel approach for discovering kinetically metastable states of biomolecular conformations. Several kinetically-aware metrics which encode both geometric and kinetic information about biomolecules are proposed. We embed the new metrics into k-center clustering and r-cover clustering algorithms to estimate the metastable states. Those clustering algorithms using kinetically-aware metrics are tested on a large scale biomolecule conformation dataset. It turns out that our algorithms are able to identify the kinetic meaningful clusters.