Chemistry Reference
In-Depth Information
completed (Zhu, Lopes, and Mackerell 2012). We have not investigated the proper-
ties of these FFs here.
5.4
SIMULATED BINARY MIXTURE KIRKWOOD-BUFF
INTEGRALS FOR POPULAR BIOMOLECULAR
FORCE FIELDS: A CASE STUDY
Our previous studies have investigated a range of FFs for simple cosolvents that
often interact with peptides and proteins. Here, we ask the question: how well do
current biomolecular FFs reproduce the binary solution properties for small solutes
representative of amino acid sidechains? Biomolecular force fields are designed
such that a protein is simply a sum of its parts, that is, small solutes can be studied
to develop the parameters required for description of functional groups commonly
found in proteins. Hence, parameters are assumed to be additive and transferable.
To this end, we present the results of a case study showing the KBIs for four binary
mixtures that were each simulated using some of the most commonly used biomo-
lecular FFs—AMBER99sb (Hornak et al. 2006), CHARMM27 (Mackerell, Feig,
and Brooks 2004; Bjelkmar et al. 2010), OPLS-AA (Jorgensen and Tirado-Rives
1988, 2005; Kaminski et al. 2001), and GROMOS54a7 (Schmid et al. 2011). Thus,
for a given FF we will interpret the results from these mixtures as a general indicator
of how successful the FF may be at reproducing solute-solute, solute-solvent, and
solvent-solvent distributions, which presumably relates to the quality of the under-
lying interaction potentials.
It is worth noting when the most recent nonbonded updates were published for
each of these FFs. The AMBER99sb release only involved updates to the torsional
parameters of AMBER99ff (Hornak et al. 2006). Among other changes, AMBER99ff
(Wang, Cieplak, and Kollman 2000), based upon AMBER94ff (Cornell et al. 1995),
did include updates to the partial charges and the addition of a few new atom types.
CHARMM's nonbonded parameters were most recently updated with the release
of the CHARMM22 protein FF (Mackerell et al. 1998). GROMOS modified their
N-H, C
=
O repulsion with the release of GROMOS 54A7 (Schmid et al. 2011) and,
prior to that, refined their nonbonded parameters for oxygen-containing functional
groups with the release of the GROMOS 53A6
OXY
parameter set (Horta et al. 2011).
Following the extensive nonbonded OPLS-AA parameterization for all amino acids
(Jorgensen and Tirado-Rives 1988), the cysteine and methionine nonbonded param-
eters were updated in the latest OPLS-AA protein FF (Kaminski et al. 2001).
The systems studied here were methanol (MOH) + water (HOH) at
x
MOH
= 0.5 at
300 K, benzene (Ben) + MOH at
x
Ben
= 0.5 and
x
Ben
= 0.75 at 308 K,
N
- methylacetamide
(NMA) + HOH at
x
NMA
= 0.1 and
x
NMA
= 0.2 at 313 K, and 1
m
and 3
m
aqueous zwit-
terionic glycine (Gly) at 300 K. The aqueous simulations were performed with the
appropriate water models: TIP3P (Jorgensen et al. 1983) for Amber, Charmm, and
OPLS; SPC (Berendsen et al. 1981) for Gromos; and SPC/E (Berendsen, Grigera,
and Straatsma 1987) for the KBFF models. The MOH + HOH system was chosen as
an example of a solution containing two polar molecules. Benzene is representative
of aromatic amino acid sidechains, and mixtures of Ben + MOH display interesting
