Free‐energy calculations of residue mutations in a tripeptide using various methods to overcome inefficient sampling

Abstract

Previous free‐energy calculations have shown that the seemingly simple transformation of the tripeptide KXK to KGK in water holds some unobvious challenges concerning the convergence of the forward and backward thermodynamic integration processes (i.e., hysteresis). In the current study, the central residue X was either alanine, serine, glutamic acid, lysine, phenylalanine, or tyrosine. Interestingly, the transformation from alanine to glycine yielded the highest hysteresis in relation to the extent of the chemical change of the side chain. The reason for that could be attributed to poor sampling of φ2/ψ2 dihedral angles along the transformation. Altering the nature of alanine's Cβ atom drastically improved the sampling and at the same time led to the identification of high energy barriers as cause for it. Consequently, simple strategies to overcome these barriers are to increase simulation time (computationally expensive) or to use enhanced sampling techniques such as Hamiltonian replica exchange molecular dynamics and one‐step perturbation. © 2016 The Authors. Journal of Computational Chemistry Published by Wiley Periodicals, Inc.

Comparison of free‐energy methods using a tripeptide‐water model system

“…The resulting moiety effectively interacts as KGK. The increased conformational freedom of this state already poses some challenges to the free‐energy methods, as described in our earlier work …”

“…After an additional 20 ns of equilibration time, we conducted a short perturbation to remove the serine side chain. This dummy side chain was kept throughout to ensure comparability with previous results and allow for future investigations. The closest water molecule adjacent to the C β was chosen to be used for the upcoming perturbations in this work.…”

“…A soft‐core potential was used to avoid singularities in forces and λ ‐derivatives resulting from overlapping atoms during an alchemical transformation . We employed the commonly used softness parameter of α VdW = 0.5 and α CRF = 0.5 nm 2 for perturbing the solute water molecule, but α VdW = 1.0 and α CRF = 1.0 nm 2 for perturbing the peptide, as described earlier . In EDS‐TI, this softness parameter was used only for the peptide perturbed by TI.…”

“…In EDS‐TI, this softness parameter was used only for the peptide perturbed by TI. While it is more efficient to employ a lower softness due to a reduced curvature of the 〈∂H/∂ λ 〉 λ curve (necessitating fewer λ ‐points), we have found previously that an increased softness can be a simple and effective tool to alleviate sampling problems in challenging systems involving end states with unequal conformational flexibility . Integrated free‐energy differences between the states remain unaffected …”

“…As the TI‐profiles of the investigated perturbations were roughly known beforehand, the distribution of simulated points along the perturbation coordinate λ could be pre‐optimized accordingly. Each perturbation was run using at least 10 and at most 14 λ ‐values (see Table S6 in the Supporting Information) spaced manually to cluster in areas of high curvature.…”