Using enveloping distribution sampling to compute the free enthalpy difference between right- and left-handed helices of a β-peptide in solution.

Recently, the method of enveloping distribution sampling (EDS) to efficiently obtain free enthalpy differences between different molecular systems from a single simulation has been generalized to compute free enthalpy differences between different conformations of a system [Z. X. Lin, H. Y. Liu, S. Riniker, and W. F. van Gunsteren, J. Chem. Theory Comput. 7, 3884 (2011)]. However, the efficiency of EDS in this case is hampered if the parts of the conformational space relevant to the two end states or conformations are far apart and the conformational diffusion from one state to the other is slow. This leads to slow convergence of the EDS parameter values and free enthalpy differences. In the present work, we apply the EDS methodology to a challenging case, i.e., to calculate the free enthalpy difference between a right-handed 2.7(10∕12)-helix and a left-handed 3(14)-helix of a hexa-β-peptide in solution from a single simulation. No transition between the two helices was detected in a standard EDS parameter update simulation, thus enhanced sampling techniques had to be applied, which included adiabatic decoupling (AD) of solute and solvent motions in combination with increasing the solute temperature, and lowering the shear viscosity of the solvent. AD was found to be unsuitable to enhance the sampling of the solute conformations in the EDS parameter update simulations. Lowering the solvent shear viscosity turned out to be useful during EDS parameter update simulations, i.e., it did speed up the conformational diffusion of the solute, more transitions between the two helices were observed. This came at the cost of more CPU time spent due to the shorter time step needed for simulations with the lower solvent shear viscosity. Using an improved EDS parameter update scheme, parameter convergence was five-fold enhanced. The resulting free enthalpy difference between the two helices calculated from EDS agrees well with the result obtained through direct counting from a long MD simulation, while the EDS technique significantly enhances the sampling of both helices over non-helical conformations.