(272a) Nonlinear Manifold Learning of Nucleosome Dynamics from Molecular Simulation

The identification of important dynamical motions remains a challenge in molecular simulations of complex systems. Here, we use a nonlinear manifold learning technique known as the diffusion map to extract key dynamical modes from molecular simulations of the nucleosome, a complex biomolecular system consisting of 147 base pairs of DNA wrapped around a disc-shaped histone octamer. We demonstrate that diffusion maps are effective at extracting motions consistent with those previously found through a detailed free energy analysis (e.g. DNA translocation along the histone complex). We also show that diffusion maps can reveal more subtle features, including the formation of looping conformations, in which DNA bulges away from the histone complex, as well as motions involved in DNA breathing, where DNA spontaneously unwraps from the histone octamer.