Chemistry Reference
In-Depth Information
1990) and found that, while both FFs were able to acceptably reproduce the enthal-
pies, densities, and diffusion coefficients of the mixtures, neither of the models could
reproduce the solution KBIs. Indeed, both models phase separated in the acetone
mole fraction range of ~0.3-0.7 (Perera and Sokolic 2004). Furthermore, alternative
force fields for urea have been observed to exhibit different quantitative behavior
concerning their interactions with small peptides (Horinek and Netz 2011).
One of the main advantages of studying solutions using KB theory is the access
to chemical potentials, in the form of their composition derivatives. We note that
instead of using KB theory, thermodynamic integration or particle insertion meth-
ods could be used to calculate the chemical potentials (Kokubo et al. 2007; Joung
and Cheatham 2008, 2009). However, these methods often do not have the precision
required to detect the typically small changes in the chemical potential with com-
position, which can often be less than a kJ/mol, without a significant computational
investment. Additionally, we emphasize that the chemical potential derivatives with
respect to composition, not the chemical potentials themselves, are directly related
to the preferential interactions generally used to quantify cosolvent effects on bio-
molecular systems (Record, Zhang, and Anderson 1998; Timasheff 1998a). Finally,
while infinite dilution free energies or enthalpies of solvation probe solute-solvent
interactions, they do not provide a check of solute-solute interactions. Thus, we pre-
fer to use KB theory as a guide, because it provides a framework to determine the
chemical potential changes for solutes of any size throughout a range of compositions.
Current FFs typically have very similar bonded terms. The vdW parameters are
usually developed to reproduce the density and enthalpy of vaporization of pure
liquids and the crystal structure dimensions. Most of the variability and uncertainty
lies in the Coulomb term. Usually, gas phase charges from
ab initio
calculations
of the geometries and minimum interaction energies between dimers are scaled to
mimic condensed phase charges, and are then tested to ensure that they reproduce
the experimental data for pure liquids. Alternatively, condensed phase partial charge
distributions are simply adopted from gas phase values that were calculated using
a basis set that created erroneously large charge distributions. We have argued that
the primary reason that many current force fields struggle to reproduce the experi-
mental KBIs lies in the approximate nature of the effective charge distributions used
to describe solute polarity, that is, the unknown degree of polarization of a solute
in a polar solvent (Weerasinghe and Smith 2003a). Consequently, instead of seek-
ing effective pair potentials that reproduce the KBIs, one could pose the legitimate
argument that explicitly polarizable FFs should naturally reproduce the KBIs bet-
ter than any nonpolarizable FF. We agree with this argument in theory. In prac-
tice, the increased computational demand of explicit polarization may prohibit those
investigators seeking to reach long simulation timescales (a significant subset of the
community) from adopting polarizable FFs (Freddolino et al. 2010). This would be
especially true for those interested in implicit solvent simulations where explicit
polarization is more difficult to include. Polarizable FFs are currently available and
examples include Ponder's Atomic Multipole Optimized Energetics for Biomolecular
Applications (AMOEBA) FF, which is currently only available for organic molecules
(Ren, Wu, and Ponder 2011), and the Chemistry at HARvard Molecular Mechanics
(CHARMM) polarizable FF for proteins, nucleic acids, and lipids, which is nearly
