Berlin 2015 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

BP: Fachverband Biologische Physik

BP 17: Posters: Protein structure and dynamics

BP 17.13: Poster

Monday, March 16, 2015, 17:30–19:30, Poster A

Langevin Modeling of Biomolecular Dynamics — •Björn Bastian and Gerhard Stock — University of Freiburg, 79104 Freiburg, Germany

Total simulation times long enough to capture biologically relevant functions are often inaccessible by Molecular Dynamics (MD) simulations on the level of full atom Newton equations. The data driven Langevin equation (dLE) technique allows for efficient propagation of a few selected system coordinates up to long simulation times on the basis of many short continuous MD trajectories (that can be computed in parallel) [1]. Beforehand, important system coordinates to describe conformational dynamics are obtained by dimensionality reduction, e.g. by principal component analysis.

If the timescales of slow system and fast bath variables separate, a general nonlinear Langevin equation can be derived from a microscopic Hamiltonian by projection techniques. The dLE algorithm presented obtains the drift, friction and diffusion fields by a local estimation on MD data. Thus dLE trajectories yield a correct global energy landscape without the requirement of input data being correctly Boltzmann weighted. Here, we present an algorithm that can treat full second-order Langevin equations in several dimensions due to more stable estimators. As proof of principle, we demonstrate the recovery of the stochastic fields for a test model.

[1] Schaudinnus N, Rzepiela AJ, Hegger R, Stock G. Data driven Langevin modeling of biomolecular dynamics. J. Chem. Phys. 138, 204106 (2013).