How robust are protein folding simulations with respect to force field parameterization?

Abstract

Molecular dynamics simulations hold the promise of providing an atomic-level description of protein folding that cannot easily be obtained from experiments. Here, we examine the extent to which the molecular mechanics force field used in such simulations might influence the observed folding pathways. To that end, we performed equilibrium simulations of a fast-folding variant of the villin headpiece using four different force fields. In each simulation, we observed a large number of transitions between the unfolded and folded states, and in all four cases, both the rate of folding and the structure of the native state were in good agreement with experiments. We found, however, that the folding mechanism and the properties of the unfolded state depend substantially on the choice of force field. We thus conclude that although it is important to match a single, experimentally determined structure and folding rate, this does not ensure that a given simulation will provide a unique and correct description of the full free-energy surface and the mechanism of folding.

Figure 1

Reversible folding simulation of the villin headpiece. The Cα-RMSD (excluding the first two and last two residues) with respect to the PDB structure 2F4K is shown for the first 100 μs of a simulation performed with the CHARMM22^∗ force field. Corresponding plots for the ff03, ff99SB^∗-ILDN, and CHARMM27 force fields are shown in Fig. S1.

Figure 2

Order of helix formation during villin folding. The order of helix formation was calculated for each folding and unfolding transition. Unfolding events were analyzed in reverse, and thus the order reported here corresponds to a folding transition. In this analysis we assign each transition a three-number code; for example, the designation 123 means that helix 1 forms first, helix 2 second, and helix 3 last.