Extracting equilibrium and dynamic properties from molecular dynamics simulations

Using molecular dynamics (MD) simulations to compute equilibrium and dynamic properties of complex systems has become a popular approach in recent years. The motivation behind these simulations is to be able to extract experimentally relevant observables, such as the folding rate of a protein, or the equilibrium population of the number of alpha-helices. I will give an introduction to Markov state models (MSM), an analysis tool for MD simulations that allows the extraction of the aforementioned equilibrium populations and transition rates from simulation data [1]. In this context its usefulness as a tool for future force field improvements taking dynamic properties into account will be discussed. This will be achieved by analysing MSMs constructed from MD data of the same model systems (peptides), but simulated using different underlying MD force fields, resulting in a direct qualative comparision of the dynamics generated by commonly used force fields. Secondly, I will give an introduction how ideas behind MSMs in combination with multiple ensemble simulations can be used to derive a new class of equilibrium estimators -- the transition based reweighting analysis method (TRAM) -- that allow the estimation of equilibrium properties from simulations, carried out using enhanced sampling techniques such as umbrella sampling or temperature replica exchange simulations. I will discussed the expanded TRAM (xTRAM) estimator using exemplary systems of varying complexity, to demonstrate improved convergence (ranging from a two fold improvement to several orders of magnitude) over commonly used estimators, such as the multistate Bennett acceptance ratio (MBAR) [2][3].

[1 ]Prinz, J.-H. et al (2011) J.Chem. Phys. 134, p.174105 [2] A.S.J.S Mey et al (2014) ArXiv:1407.0138 [3] M.R. Shirts and J.D. Chodera (2008) J. Chem. Phys. 129 p. 124105