GNEIMO: Constrained Molecular Dynamics Methods For Long Time Scale Simulation of Macromolecules

The two leading causes for the limitation of the all-atom molecular dynamics (MD) simulation timescales are: 1. calculation of forces that scale as the square of the number of atoms and, 2. the integration time step is limited to 1fs due to the high frequency modes in the protein. High performance technologies and better force calculation algorithms have addressed the former, and we address the latter issue in this work.Here we report a constrained MD method, GNEIMO (Generalized Newton-Euler inverse mass operator method), that is capable of achieving stable dynamics with integration time steps as large as 10 to 20fs. The GNEIMO method provides a platform to perform long time scale hierarchical simulations ranging from all-atom simulations, coarse-grained dynamics of clusters of few atoms, to dynamics with larger motifs constrained, at lesser computational expense compared to all-atom MD. GNEIMO method uses spatial operator algebra to solve for the internal coordinate dynamics with computational cost scaling linearly as the number of degrees of freedom.The current implementation of GNEIMO is capable of performing constant temperature Nose-Hoover dynamics, with continuum Generalized Born solvation. We use adaptive step size integration to provide stable dynamics with larger time steps. We have carried out tens of nano-seconds of stable dynamics for proteins with time steps as large as 10 to 20fs for different integrators. We report results from conformational changes of domains in proteins from long time scale dynamics. This implementation integrates the force-field module from the MD program LAMMPS, with constrained dynamics module from NASA-Jet Propulsion laboratory.