Comparing the Efficiency of Biased and Unbiased Molecular Dynamics in Reconstructing the Free Energy Landscape of Met-Enkephalin

Abstract

All-atom unbiased molecular dynamics simulations are now able to explore the microsecond to millisecond time scale for simple biological macromolecules in an explicit solvent. This allows for a careful comparison of the efficiency and accuracy of enhanced sampling methods versus long unbiased molecular dynamics in reconstructing conformational free energy surfaces. Here, we use an equilibrium microsecond-long molecular dynamics simulation as a reference to analyze the convergence properties of well-tempered metadynamics with two different sets of collective variables. In the case of the small and very diffusive Met-enkephalin pentapeptide, we find that the performance strongly depends on the choice of the collective variables (CVs). Using a set of principal component analysis derived eigenvectors, the convergence of the FES is faster than with both hand-picked CVs and unbiased molecular dynamics.