An Improved OPLSAA Force Field for Carbohydrates

D. KONY,1 W. DAMM,2 S. STOLL,1 W. F. VAN GUNSTEREN2

1

CABE, Department of Inorganic, Analytical and Applied Chemistry, Science II, University of
Geneva, 30 Quai E. Ansermet, CH-1211 Geneva 4, Switzerland
2
Laboratory of Physical Chemistry, ETH Zurich, ETH Honggerberg,
CH-8093 Zurich, Switzerland

Received 6 September 2001; Accepted 16 May 2002

Published online XX XXXX 2002 in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/jcc.10139

Abstract: This work describes an improved version of the original OPLSall atom (OPLSAA) force field for
carbohydrates (Damm et al., J Comp Chem 1997, 18, 1955). The improvement is achieved by applying additional
scaling factors for the electrostatic interactions between 1,5- and 1,6-interactions. This new model is tested first for
improving the conformational energetics of 1,2-ethanediol, the smallest polyol. With a 1,5-scaling factor of 1.25 the
force field calculated relative energies are in excellent agreement with the ab initio-derived data. Applying the new
1,5-scaling makes it also necessary to use a 1,6-scaling factor for the interactions between the C4 and C6 atoms in
hexopyranoses. After torsional parameter fitting, this improves the conformational energetics in comparison to the
OPLSAA force field. The set of hexopyranoses included in the torsional parameter derivation consists of the two
anomers of D-glucose, D-mannose, and D-galactose, as well as of the methyl-pyranosides of D-glucose, D-mannose.
Rotational profiles for the rotation of the exocyclic group and of different hydroxyl groups are also compared for the
two force fields and at the ab initio level of theory. The new force field reduces the overly high barriers calculated using
the OPLSAA force field. This leads to better sampling, which was shown to produce more realistic conformational
behavior for hexopyranoses in liquid simulation. From 10-ns molecular dynamics (MD) simulations of -D-glucose and
-D-galactose the ratios for the three different conformations of the hydroxymethylene group and the average 3JH,H
coupling constants are derived and compared to experimental values. The results obtained for OPLSAASEI force field
are in good agreement with experiment whereas the properties derived for the OPLSAA force field suffer from
sampling problems. The undertaken investigations show that the newly derived OPLSAASEI force field will allow
simulating larger carbohydrates or polysaccharides with improved sampling of the hydroxyl groups.
2002 Wiley Periodicals, Inc.

J Comput Chem 23: 1416 1429, 2002

Key words: carbohydrates; 1,2-ethanediol; hexopyranoses; conformational analysis; force field; molecular dynamics

Introduction
Polysaccharides as well as even simple monosaccharides are challenging molecules to model due to their numerous highly flexible
and polar OH groups. By forming intra- and intermolecular hydrogen bonds these groups determine the conformational space
sampled.1 Using a simple force field with atom centered charges is
a necessary compromise in accuracy for performing time consuming molecular dynamics (MD) studies in which the solvent is
treated explicitly.2
The force fields available to study biomolecular systems consist
of very similar potential energy functions, and are converging to
the same approach regarding the derivation of force field parameters. The nonbonded parameters are adjusted to reproduce the
physical properties of pure liquids such as the heat of vaporization
and the density. Following this approach, it is necessary to represent the polarized liquid state in an average way. This is achieved

by using 10 20% larger charges than necessary to reproduce the

gas phase molecular dipole moments.3,4 The torsional parameters
are fit to reproduce the gas phase conformational energetics, usually derived with ab initio methods, of fragments or model systems
representing the chemical environment of the torsional angle under
consideration. This allows incorporating stereoelectronic effects in
the otherwise pure mechanical force field model. The force fields
derived with these two key elements have been found to obtain
excellent free energies of solvation and free energies of binding.5

Correspondence to: W. F. van Gunsteren; e-mail: wfvgn@igc.phys.chem.

ethz.ch
Contract/grant sponsors: Swiss National Reserach Project; contract/grant
number: 2100-0621750.00/1
Contract/grant sponsor: University of Geneva

2002 Wiley Periodicals, Inc.

Improved OPLSAA Force Field for Carbohydrates

Force field parameter developments for studying carbohydrates

in the condensed phase are rare6 15 in comparison to parameter
developments for proteins and other simple organic molecules.
The latest AMBER13 force field uses partial charges generated
from a least-square fit to an ab initio calculated electrostatic
potential, which is done for multiple D-glucose conformations
(RESP fitting).16 When only two glucose conformations are used
for the fitting a larger partial charge (O: 0.68 e, H: 0.46 e) is
obtained on oxygen atoms of hydroxyl groups, whereas when 12
conformations are used smaller charges (O: 0.54 e, H: 0.37 e) are
obtained. The charge set with the smaller charge on oxygen atoms
is then used for the derivation of torsional parameters and in liquid
simulations. The larger charge is in the order of a partial charge
used on a hydroxyl oxygen atom in the AMBER or the OPLS
AA17 force fields for simple monofunctional alcohols. The smaller
charge assigned to carbohydrates will therefore lead to smaller
interaction energies and therefore also to weaker solvation and
binding. This could lead to artifacts in studies where a carbohydrate is bound to a receptor with several monofunctional hydroxyl
groups.
The nonbonded parameters for carbohydrates of the OPLSAA
force field are derived by studying the liquid properties of 1,2ethanediol.18 With these parameters the experimental values for
the heat of vaporization and density of the pure liquid are reproduced. The charges on the hydroxyl functionality (O: 0.7 e, H:
0.435 e) are slightly larger than those of monofunctional alcohols
(O: 0.683 e, H: 0.418 e). Even larger charges had to be used to
describe the pure liquid properties of glycerol (O: 0.73 e, H:
0.465 e).19 This indicates that the longer the polyolic chain of the
monomer is, the larger the polarization effects in the pure liquid
become. The magnitude of these charges are all within the range of
those obtained from HF/6-31G* CHELPG fittings to reproduce the
electrostatic potential and therefore similar to those of the GLYCAM force field.12 The torsional parameters of the OPLSAA
force field9 are also fit to reproduce the conformational energetics
derived with ab initio calculations. In contrast to the latest AMBER force field13 not only D-glucose but also D-galactose and
D-mannose are used to obtain a wider applicable set of parameters.
The derivation of the torsional parameters for the CHARMM
force field14 follows a different concept. Only fragments, such as
1,2-ethanediol or hydroxymethanol, featuring the chemically different torsional environments of a hexopyranoses are used. However, it is questionable if with those torsional parameters the
conformational energetics of intact hexopyranoses can be reproduced.
Force field parameters have also been tested and refined against
crystal structure data.7,8 To successfully predict crystal structures,
the intermolecular interactions have to be in balance with the
intramolecular interactions and refined to describe the condensed
phase accurately. van Eijck and Kroon developed a program (UPACK) for the generation of crystal structures by means of energy
minimization, and have recently compared the results of three all
atom force fields with geometric and energetic criteria.8 The
OPLSAA was heralded as the most promising all-atom force
field. This indicates that the energetics described with the OPLSAA
force field is already rather refined. Unfortunately, these investigations do not give any information about the dynamical behavior.

1417

Figure 1. Conformational energies for 1,2-ethanediol and -D-galactose for which the largest deviation between the OPLSAA force field
and the ab initio level of theory is obtained.

Despite the promising results obtained for the OPLSAA force

field so far, large discrepancies in reproducing the gas phase
relative energies are found as well. The relative energies of the two
conformations with which the exo-anomeric effect is represented,
cannot be reproduced (Fig. 1).9 A large energy discrepancy of
about 4 kcal/mol is also observed when the two 1,2-ethanediol
conformations are considered which represent the same difference
in geometry. One factor contributing to this discrepancy originates
from the fact that the electrostatic 1,4-interactions are reduced by
a factor two, which causes an imbalance with the 1,5-interactions
(Fig. 1). The discrepancy between the force field and the ab initio
calculated relative energy is caused by the inclusion of the mean
field approximation used to describe the polarized liquid state.
With smaller charges on the hydroxyl functionality the ab initio
gas phase energetics could be much better reproduced, whereas
with the larger charges as derived for glycerol the discrepancy is
increased. The too high energy of nonhydrogen bonded structures
as shown in Figure 1 for 1,2-ethanediol lead in hexopyranoses to
very high barriers for rotation of the hydroxyl group, which leads
to sampling problems in Monte Carlo simulations using explicit
solvent.20 Such sampling problems make it impossible to accurately study the conformational behavior of carbohydrates.
The OPLSAA force field has also been used to study the
conformational space of -D-Manp-(1-3)--D-Glcp-Ome disaccharide in explicit water and DMSO.21 The same systems were
also run with the GLYCAM force field for carbohydrates for
which similar results are obtained. Surprisingly, no sampling problems are reported using the OPLSAA force field. Rather, fast
oscillations between g and g conformations with respect to the
HCOH torsions are observed.

1418

Kony et al.

Vol. 23, No. 15

In this article we describe the development of a new version of

the OPLSAA force field for carbohydrates. We are aiming to
improve the largest discrepancies of the conformational energetics
obtained for the OPLSAA force field and to better reproduce the
high energy conformations with broken intramolecular hydrogen
bonds which are populated most in aqueous solution. This is done
by applying an additional scaling factor for the 1,5- and specific
1,6-interactions, which reduce the strength of the electrostatic
interactions in hexopyranoses and 1,2-ethanediol. Because these
additions change the strength of the intramolecular nonbonded
interaction, the torsional parameters are reoptimized. The resulting
parameter set is then tested in the liquid phase for obtaining better
sampling and for reproducing experimentally derived 3JH,H coupling constants. With an improved sampling obtained for hexopyranoses, larger carbohydrates could be studied by simulating for
several nano- or microseconds in explicit solvent as has been
successfully done for peptides22 and proteins.23 The liquid phase
simulations reported here demonstrate that the improved gas phase
energetics of the newly developed force field leads to a more
realistic solvation and conformational behavior of hexopyranoses
in aqueous solution.

Computational Methods
Ab initio and Density Functional Theory (DFT)
Calculations

Ab initio and DFT calculations are performed using the GAUSSIAN94 program.24 For 1,2-ethanediol, all structures are optimized
at the B3LYP/6-311G** level of theory. It has been found for
this molecule that B3LYP/6-311G** gives similar results to those
obtained using a very high level of quantum theory.25,26 Introduction of a diffuse function into the basis set, which significantly
reduces the basis set superposition error (BSSE) in the hydrogen
bonding energy calculation, leads to better potential energy surfaces of carbohydrates.1 Thus, the 6-311G** basis set is chosen
for our work. Ten conformations are reported to represent true
minima for 1,2-ethanediol.27 For the torsional parameter fitting,
further rotamers are generated. For the conformations with a
OCCO torsion () gauche and trans, rotamers in 30 increments of the CCOH torsional angle 1 are generated, whereas
the angle 2 is fixed at 180 (Fig. 2).
For hexopyranose, the many conformations are available from
the work of Damm et al.,9 which are obtained via full geometry
optimizations using RHF theory and the 6-31G* basis set. The
HF-6-31G* level of theory is known to be adequate for glucose
conformations and can only be improved at an unattainable com-

Figure 2. Dihedral angle definition and partial charges (in e) used in

the force field calculations for the 1,2-ethanediol.

Journal of Computational Chemistry

Figure 3. Pairs of atoms describing 1,6- and 1,5-interactions for

which additional scaling factors are applied in the OPLSAASEI
force field. For the 1,5 pairs, an example of the application of the
scaling factor is given for one of the 1,5 pairs of atoms for hexopyranose (1) and glycol (2).

putational expense considering the number of conformations

used.28 Our additional electronic structure calculations for deriving
further hexopyranose conformations are therefore also performed
at the RHF/6-31G* level, which has been used previously for
developing torsional parameters.29
Force Field Studies

The OPLSAA potential energy functions and parameters are

used9 with the exception of the torsional parameters, which are
newly derived, and the additional scaling factors used. The 1,5electrostatic interactions are reduced for 1,2-ethanediol and for
hexopyranoses by a factor 1.25 and 1.26, respectively. In addition,
for hexopyranoses a scale factor equal to 1.22 is applied between
the 1,6-interacting atoms of the hydroxymethylene and the vicinal
hydroxyl groups (Fig. 3). This scaling factor is necessary to restore
the balance with the 1,5-interactions.
Molecular mechanics calculations are performed with the
GROMOS96 package30 that was adapted to reduce the 1,4-, 1,5-,
and 1,6-electrostatic interactions. For geometry optimizations in
GROMOS96, the conjugate-gradient algorithm is used with a
convergence criterion of 1 105 kJ/mol. The POTENT program31 is modified in order to provide the molecular topology file
with the OPLSAA parameters for the GROMOS96 program. In
addition to the standard OPLSAA parameter data file, only
atomic coordinates and OPLS atom types are required for the input
file.
The following conditions are used for each simulation. Each
system consists of a single -D-hexopyranose surrounded by water
in a cubic box subject to periodic boundary conditions to eliminate
edge effects. For water the TIP3P model32 is used. The shortest
distance between the hexopyranose and the wall of the box is
initially 1.2 nm, where the number of water molecules varied from
580 to 600, depending on the hexopyranose molecule and the
initial conformation. The simulations are performed at 1-atm pressure and 300 K by weak coupling33 to an external bath using the
coupling constants of 0.1 ps for the temperature and 0.5 ps for the
pressure coupling. A cutoff radius of 9 for nonbonded interac-

Improved OPLSAA Force Field for Carbohydrates

tions is used and a nonbonded pair list is updated every 5 MD

steps. After a minimization of the entire system, a 100-ps MD
simulation is performed with restraining the solute degrees of
freedom to relax the surrounding solvent molecule, and to attain a
well solvated equilibrium system. Following this, the sampling
period or analysis period is performed and configurations are
stored every 0.5 ps. Averages for 3JHH coupling constants are
calculated using a generalized Karplus equation:34
JH,H 13.22 cos2 0.99 cos

xi 0.87 2.46 cos2 i19.9xi.

In this equation denotes the protonproton torsion, i stands for
1 or 1 according to the orientation of the substituent with
respect to its geminal proton. The factors of electronegativity (xi)
are set 1.3 (O5 and O6) and 0.4 (C4).
Mean values for torsional angles of the hydroxymethylene
group are determined according to the following procedure. A
conformation is considered g if the torsion value is between 0 and
120 degrees, t if the value is between 120 and 240 degrees, g if
the value is between 240 and 360 degrees.
Torsional Parameter Optimization

The FITGR program (a GROMOS adapted version of FITPAR

program35) is used to derive the torsional parameters. The heart of
the FITGR routine is identical to the FITPAR routine. A difference
function between the ab initio and molecular mechanics energies is
constructed among a set of structures and then minimized by
torsional parameters variation using the combination of a simplex
and FletcherPowell algorithm. In contrast to FITPAR, FITGR
reads the GROMOS topology and the two GROMOS geometry
optimization output files containing the atomic coordinates and
potential energy. The introduction of the new scale factor for the
1,5- and 1,6-electrostatic interaction changes the nonbonded intramolecular potential. Following the procedure described previously9 the set of torsional parameters optimized for the OPLSAA
force field is adjusted. This set of torsional parameters is augmented by additional types (see below).

Results and Discussion

1,2-Ethanediol

In the following section we discuss the development of the OPLS

AASEI (OPLSAA Scaling Electrostatic Interaction) force field
and compare the gas phase energetics obtained from the ab initio
and force field calculations using 1,2-ethanediol. In Table 1 the
relative energies and the torsional angles (, 1, 2 see Fig. 2)
obtained for the 10 possible minima and the calculated rotamers
are given. The relative energies obtained for the 10 minima at the
B3LYP/6-311G** level of theory are very close to the MP2
energies obtained by Cramer and Truhlar,27 which confirms the
quality of the additional rotamers.
The OPLSAA and newly optimized OPLSAASEI torsion
parameters are given in Table 2. The torsional parameters describ-

1419

ing the OCCO torsion with a single V1 term did not optimally
describe the energetics of 1,2-ethanediol with the OPLSAASEI
electrostatic model. Thus, a set of three Fourier coefficients has
been used in the fitting.
The relative energies for the 1,2-ethanediol conformers and
rotamers are calculated using the standard OPLSAA parameters,
the OPLSAASEI parameters and at ab initio level. The results
are given in Table 1 and depicted in the rotational profile of Figure
4. The RMSD between the ab initio and the force field calculated
relative energies for all conformers is 0.07 and 0.51 kcal/mol for
the OPLSAASEI and the standard OPLSAA, respectively. The
range of energies spanned by the conformers is ca. 5.5 kcal/mol for
the ab initio and the OPLSAASEI models and about 10 kcal/mol
for the standard OPLSAA. The energy of the intramolecular
hydrogen bond between the two hydroxyls is evaluated by the
energy difference between the tGg and tGt conformers (notation explained in ref. 27). A large deviation between the ab initio
and OPLSAA calculations is found for the relative energies of
these two conformers, as shown in Figure 1. In contrast to the
OPLSAA force field, the OPLSAASEI force field correctly
reproduces the ab initio results for these two structures (Table 1).
The relative energy between a trans and a gauche conformation of
a CCOH dihedral angle, presented by the structures tGg and
gGg, does not involve creating or breaking a hydrogen bond.
The OPLSAASEI closely reproduces the ab initio energy for
these two structures and the energy difference of 0.5 kcal/mol
(0.30 kcal/mol for the OPLSAA) between them. Structures tGt
and tTt allow us to compare the reproduction of the G,T relative
energies of the ab initio and DFT calculations (0.6 kcal/mol and
0.9 kcal/mol, respectively) by the two force fields. OPLSAASEI
performs well with a relative energy of 1.25 kcal/mol in comparison to OPLSAA, which gives a relative energy for the tGt
conformation of almost 8 kcal/mol.
The energy profiles for the rotation around the CCOH
torsion calculated at the ab initio, OPLSAA and OPLSAASEI
level are presented in Figure 4 for the G and T conformers. The
OPLSAASEI force field calculated the g and g minima
shifted by 30 with respect to the ab initio calculated values for the
T conformer. Besides this deviation, the agreement between the
quantum mechanics results and the OPLSAASEI results is very
good. The largest deviations are obtained for the OPLSAA force
field, which yields a too large barrier for the two profiles. In
particular, for the rotational profile in which the OCCO dihedral is arranged gauche, the OPLSAA force field deviates from
the ab initio calculated relative energy by more than 4 kcal/mol.
This indicates that the additional scaling factor dramatically improve the gas phase conformational energetics for 1,2-ethanediol
and that applying this scaling factor will also improve the conformational energetics for carbohydrates.
Hexopyranoses
Torsional Parameter Optimization

The torsional parameters assigned to hexopyranose structures are

given in Table 3. The identical types are kept fixed during
torsional parameter optimization. As in the standard OPLS version
for carbohydrates, these parameters are taken from OPLSAA

1420

Kony et al.

Vol. 23, No. 15

Journal of Computational Chemistry

Table 1. Relative Energies (kcal/mol) of the Glycol Conformers, Calculated at Different Levels of Theory,

Together with the Results of the OPLSAASEI and OPLSAA Force Fields.
GLYCOL conformersa
g G g
g G g
g G g
g Tg
g Tg
tG g
tG g
tG t
tTg
tTt
transi
transi
transi
transi
transi
transi
transi
gauchei
gauchei
gauchei
gauchei
gauchei
gauchei
gauchei
gauchei
gauchei
gauchei
gauchei
gauchei

B3LYPb

MP2c

OPLSAAd

OPLSAASEIe

1()f

()f

2()f

1.18
2.96
0.49
2.69
2.99
3.76
0.00g
3.54
2.79
2.62
5.16
4.13
2.90
3.00
3.29
2.92
2.62
3.23
3.97
3.78
4.30
4.75
4.26
3.54
3.38
3.00
1.50
0.17
0.91

1.20
3.22
0.31
2.99
2.80
3.81
0.00h
3.48
2.91
2.85

1.47
4.02
0.30
2.88
1.81
8.16
0.00
7.87
2.31
1.79
4.65
3.69
2.43
2.65
3.18
2.46
1.79
4.18
7.15
8.28
9.61
10.34
9.32
8.07
7.76
6.87
4.07
0.99
0.56

1.28
3.14
0.50
3.75
2.63
4.78
0.00
3.80
2.98
2.55
5.57
4.68
3.34
3.06
3.30
2.92
2.55
3.61
4.72
4.25
4.58
4.90
4.28
3.53
3.26
2.60
1.02
0.02
1.13

277.7
45.7
75.1
65.9
69.7
183.9
193.4
180.0
186.5
180.0
0.0j
30.0j
60.0j
90.0j
120.0j
150.0j
180.0j
0.0j
30.0j
60.0j
90.0j
120.0j
150.0j
180.0j
210.0j
240.0j
270.0j
300.0j
330.0j

60.8
53.5
58.1
176.6
179.9
66.2
61.7
60.0
180.2
180.0
180.0
182.4
180.1
177.6
176.2
176.5
180.0
54.7
62.7
66.3
69.0
71.0
72.4
74.5
75.6
74.1
70.0
63.3
55.5

277.5
45.9
314.7
66.0
290.4
56.5
306.6
180.0
70.8
180.0
180.0j
180.0j
180.0j
180.0j
180.0j
180.0j
180.0j
180.0j
180.0j
180.0j
180.0j
180.0j
180.0j
180.0j
180.0j
180.0j
180.0j
180.0j
180.0j

0.51

0.07

RMSDk

The label for each conformation indicates the 1, , 2, dihedral angle. t (or T) means trans, g (or G ) means
gauche and g means a gauche conformation.
b
B3LYP/6-311 G** relative energies in kcal/mol.
c
MP2/cc-pVTZ//MP2/cc-pVTZ relative energies in kcal/mol (ref. 27).
d
Relative energies from the OPLSAA force field in kcal/mol.
e
Relative energies from the OPLSAA force field in kcal/mol with a torsional parameter set optimized using a scale
factor for the 1,5 electrostatic interactions equal to 1.25.
f
B3LYP/6-311G** optimized geometry.
g
Energy 230.3295503 au.
h
Energy 229.86044 au (ref. 27).
i
Conformation is shown in rotational profile of Figure 4.
j
Fixed dihedral angle.
k
RMSD for the energy difference between the B3LYP/6-311G** and the force field calculated relative energies of the
corresponding column.
a

values for alkanes, alcohols and ethers17 where scaling the 1,5electrostatic interactions does not require a parameter refinement.
The new type are those optimized for the OPLSAASEI force
field. For the OHCCOH and CCOH torsion the same analytical form as for 1,2-ethanediol is taken for the hexopyranoses.
However, it turns out that the 1,2-ethanediol Fourier coefficient
values can be further improved for hexopyranose and they are

therefore reoptimized. The three sets of Fourier coefficients are

optimized for the OCO(H, C) torsional types. In contrast to the
OHCCOH type, we find during torsional optimization that a
single V1 term is sufficient for the OSCCOH torsion. The
OPLSAASEI force field has three CCCO parameters sets:
CCCOH for dihedral angle to the hydroxyl oxygen, CCCOS
for the dihedral angle to the ether oxygen, and CCCOS_ace that

Improved OPLSAA Force Field for Carbohydrates

1421

Table 2. Sets of Torsional Angle Parameters for Glycol for the OPLSAA Force Field and for the

OPLSAA Force Field using a 1,5 Scale between Electrostatics Interactions Corresponding to the
Potential Energy Function.a
Etorsion

Vi1
1 cos f1
2

Vi2
1 cos2 f2
2

Vi3
1 cos3 f3
2

OCCO

OPLSAA
OPLSAASEIc

CCOH

V 1b

V 2b

V 3b

V 1b

V 2b

V 3b

9.508
4.780

0.000
3.593

0.000
0.887

0.356
0.004

0.174
0.629

0.492
0.035

is the dihedral angle, and f 1 , f 2 , f 3 are all zero.

Constants are given in kcal/mol. The HCOH, HCCO, HCCH parameters are not optimized and stay identical to the
standard OPLSAA values (ref. 9).
c
The scale factor for the 1,5 electrostatic interactions is equal to 1.25.
a

relates to the dihedral angle between the C3 ring carbon and the
acetal group positioned on the C1 ring carbon. As for the unique
CCCO type of OPLSAA, a single V1 term is sufficient for the
CCCOH. In contrast, three terms are more suitable for the
CCCOS torsion type. To correctly reproduce the methyl-hexopyranoside relative energies, it is necessary to add the new
CCCOS_ace type. Ace means acetal and indicates the position of this dihedral in the molecule.

Figure 4. Profiles for rotation around the CCOH torsion in the

T-(a) and G-(b) 1,2-ethanediol conformation calculated ab initio at
B3LYP/6-311G** level and with the OPLSAASEI and OPLSAA force fields.

Comparison of Relative Energies Obtained with the Two

OPLSAA Force Fields and at Ab Initio Level of Theory

In the following the performance of the OPLSAA and OPLS

AASEI force fields in reproducing ab initio calculated relative
energies of five different hexopyranoses is compared. The 44
hexopyranoses structures included in the OPLSAA parameter
development are evaluated.9 Their relative energies are given in
Table 4. The root-mean-square deviation (RMSD) between the
OPLSAASEI (OPLSAA) and the ab initio calculated relative
energy is 0.61 (0.75) kcal/mol for all the 44 conformers. The
RMSD for the 33 hexopyranoses with a hydroxyl at C1 (1, 2, 3;
Fig. 5) is 0.65 (0.83) kcal/mol and 0.57 (0.42) kcal/mol for the 11
hexopyranosides with a methoxy group at C1 (4, 5; Fig. 5). The
similar RMSD obtained for the OPLSAASEI and the OPLSAA
force field indicates that the new scheme with the reduced electrostatic interactions fits the torsional parameter with the same
quality only. This can be understood because the set of carbohydrate conformations consists of low energy conformations only. As
the comparison of the conformational energetics for glycol shows,
especially the high energy conformations are better reproduced with
the new force field so that the RMSD obtained for the 44 low-energy
carbohydrate conformations is not reflecting an improvement.
The largest discrepancy (3.8 kcal/mol) between the ab initio
and the OPLSAA force field calculated energies is obtained for
conformation 2f (Fig. 1). As expected, the OPLSAASEI force
field considerably improves the reproduction the ab initio relative
energy of 2f. This shows that the additional scaling factor between
15 interacting atoms restores the balance between the 1 4 and
15 electrostatic interactions. Consequently, the OPLSAASEI is
able to reproduce the energetics of all the hexopyranoses showing
this structural change describing the exo-anomeric effect.
The corresponding -anomers are represented by the structures
2g and 2h (Fig. 6). Both force fields yield a too high relative
energy for these two structures with respect to the ab initio level
calculations. However, the OPLSAASEI (OPLSAA) force
field gives a larger deviation for both structures 2g and 2h being
1.5 (0.61) kcal/mol and 1.2 (0.95) kcal/mol.

1422

Kony et al.

Vol. 23, No. 15

Journal of Computational Chemistry

Table 3. Sets of Torsional Angle Parameters for Hexopyranoses for the OPLSAA and OPLSAASEI

Force Fields Corresponding to the Potential Energy Function.a

Etorsion

Vi1
1 cos f1
2

Vi2
1 cos2 f2
2

Vi3
1 cos3 f3
2

OPLSAA
V 1b
Identical types
CCCC
CCCH
HCOC
COCC
HCCO
HCCH
HCOH
Modified types
OCOC
OSCCOH
CCOH
OHCCOH
OCOH
CCCOH
CCCOS
CCCOS_ace

1.740
0.000
0.000
0.650
0.000
0.000
0.000
0.375
4.319
2.674
9.066
1.257
1.336
1.336
1.336

OPLSAASEI

V 2b

V 3b

0.157
0.000
0.000
0.250
0.000
0.000
0.000

0.279
0.366
0.760
0.670
0.468
0.318
0.450

1.358
0.000
2.883
0.000
1.806
0.000c
0.000c
0.000c

0.004
0.000
1.026
0.000
0.003
0.000c
0.000c
0.000c

V 1b

V 2b

V 3b

1.740
0.000
0.000
0.650
0.000
0.000
0.000

0.157
0.000
0.000
0.250
0.000
0.000
0.000

0.279
0.366
0.760
0.670
0.468
0.318
0.450

3.687
1.593
0.205
8.519
1.673
2.587
0.089
1.501

0.005
0.000
1.459
6.829
2.200
0.000
0.729
0.000

0.537
0.000
0.010
4.793
0.001
0.000
0.870
0.000

is the dihedral angle, and f 1 , f 2 , f 3 the phase shifts. All phase shifts are zero.
Constants are given in kcal/mol.
c
CCCO is unique in OPLSAA.
a

An estimate of the energy difference between the conformers

with which the anomeric effect is described can be deduced from
the energy difference between structure 1c and its corresponding
anomer 1j. The ab initio (1.17 kcal/mol) and OPLSAA (1.22
kcal/mol) calculated energy differences are in a good agreement,
whereas the OPLSAASEI calculated energy difference is too
high (1.94 kcal/mol). Finally, the relative energy of methyl-manno-pyranoside 5c is obtained too low by the OPLSAASEI
force field by about 1.4 kcal/mol. This large discrepancy leads to
a slightly larger value for the RMSD of the 11 methyl-hexopyranoses in comparison to the one derived with the OPLSAA force
field.
These results show that despite the fact that the OPLSAASEI
force field considerably improves the largest discrepancies of the
OPLSAA force field, it does not reproduce the relative energy of
some carbohydrate characteristic structural features as accurate as
the OPLSAA force field. However, these few discrepancies could
turn out to be less important for the overall quality of the molecular
model. This is illustrated in the following by comparing rotational
profiles calculated with both force fields and at ab initio level.
Comparison of Rotational Profiles Obtained with the
OPLSAA and OPLSAASEI Force Fields and at the Ab
Initio Level of Theory

Rotational profiles for the hydroxyl group at C1: the profiles for
the rotation around the C1O1 bond are shown for the - and

-anomers of D-glucopyranose and D-mannopyranose (Fig. 7).

They describe the energetics of the exoanomeric effect in hexopyranoses and show how the epimerization at C2, going from Dglucose to D-mannose, changes the energetics of the OCOH
profile. Overall, the OPLSAASEI improves the description of
this effect in comparison to the OPLSAA force field. However,
there are still discrepancies, which are important to consider in
future work. Parallel work done includes optimization of torsional
parameters using only the glucopyranose conformers for the fitting. The results show that in this case the parameters are able to
reproduce the relative energy of rotamers and lowest energy conformers with much better accuracy with respect to the ab initio.
Therefore, considering other hexopyranoses, as the two anomers of
D-mannose in this case, into the fitting significantly increases the
complexity of system and therefore leads to a lower quality of the
OCOH rotational profile calculated for D-glucose.
Rotational profiles for the hydroxymethylene group: the potential energy profiles for rotation around the C5C6 bond are examined for the -D-glucopyranose and the -D-galactopyranose (Fig.
8). The effect of the epimerization at C4, corresponding to the
structural difference of -D-glucopyranose and -D-galactopyranose, on the energetics of the different orientations of the hydroxymethylene group is investigated.36 The hydrogen of the
hydroxyl at C6 is held fixed at 180 with respect to the C5 atom so
that the possible intramolecular hydrogen bonds between the hydroxymethylene group and the adjacent hydroxyl group and the

Table 4. Relative Energies (kcal/mol) of Hexopyranoses at Ab Initio Level of Theory, and with the OPLS
AA and OPLSAASEI Force Fields.
Conformersa
1a
1b
1c
1d
1eg
1fh
1gh
1hg
1i
1jg
1k
2a
2b
2c
2d
2eg
2fg
2gg
2hg
3a
3bg
3c
3d
3eh
3fh
3gh
3h
3i
3j
3k
3l
3mh
3nh
4a
4b
4c
4d
4e
4f
4g
4h
5a
5b
5c

/cc/gg
/cc/gg
/cc/tg
/cl/gg
/cl/tt
/cl/tt
/cl/tt
/cc/gg
/cc/gg
/cc/tg
/cc/tg
/cl/gg
/cc/gg
/cc/gg
/cc/tg
/cc/tt
/cl/gg
/cc/gg
/cc/tg
/cl/gg
/cl/gg
/cl/tt
/cc/tg
/cl/tt
/cl/tt
/cl/tt
/cc/gg
/cc/gg
/cc/tg
/cl/gg
/cl/tt
/cl/tt
/cl/tt
/cc/gg
/cc/gg
/cc/tg
/cl/gg
/cl/tt
/cc/gg
/cc/gg
/cc/tg
/cl/tt
/cc/tg
/cc/tg

HFb

B3LYPc

OPLSAASEId

OPLSAAe

OCOX ()f

0.86
0.94
0.74
1.66
1.67
2.08
1.94
1.86
2.06
1.91
2.99
0.00k
1.89
0.94
0.97
2.36
2.29
2.55
2.42
0.99
2.93
1.44
2.96
2.61
1.45
2.69
2.22
2.32
1.86
1.61
1.82
2.12
2.49
0.11
0.00n
0.00
2.90
2.94
1.23
1.46
1.29
0.99
2.49
1.61

0.57
0.58
0.57
1.48
1.71
2.18
1.81
1.31
1.20
1.42
2.34
0.00l
1.56
0.77
1.55
2.53
2.10
2.07
2.77
1.35
2.94
1.96
2.95
3.18
1.95
2.70
2.03
1.81
1.72
1.49
1.92
2.36
2.29
0.22
0.00o
0.16
3.05
3.33
1.11
1.21
1.04
1.97
2.81
1.93

0.44
0.68
0.42
1.77
1.63
1.76
1.34
2.58
2.57
2.36
2.20
0.00
2.42
0.86
0.73
1.72
2.43
4.05
3.60
1.07
3.90
1.20
2.42
2.37
1.43
1.26
1.64
1.13
1.20
2.27
1.49
1.56
2.54
0.11
0.00
0.12
3.14
3.19
1.79
1.88
1.94
1.03
1.80
0.18

1.13
0.94
0.84
1.28
1.93
2.10
2.45
2.43
2.06
2.09
2.64
0.00
2.97
0.65
0.94
2.35
6.14
3.16
3.37
1.18
2.94
1.32
2.99
2.70
1.46
1.50
2.28
0.93
1.21
1.81
1.49
1.70
2.51
0.32
0.00
0.06
2.95
3.44
2.04
1.52
1.88
0.58
2.04
1.66

63.3
61.7
62.1
164.9
164.7
178.9i
152.2j
61.6
62.9
58.5
38.7
165.5
63.6
62.1
62.7
62.1
69.1
61.1
53.0
58.5
57.2
58.1
53.3
30.0m
60.0m
90.0m
48.7
49.2
45.0
74.4
72.2
60.0m
90.0m
67.7
64.9
67.0
72.1
71.8
68.4
68.8
66.7
66.3
63.6
62.4

0.61
0.65
0.57

0.75
0.83
0.42

RMSDp
All 44 conformations
Hexopyranoses
Methyl-hexopyranosides

a
The label for each conformation indicates the anomer/the orientation of the hydrogen-bond network/the OC5C6 O and the
C5C6 OH dihedral angles. t means trans, g means gauche, a g means gauche conformation. cc stands for counterclockwise
and cl for a clockwise hydrogen bond network.
b
Relative energies obtained at RHF/6-31G*//RHF/6-31G* in kcal/mol (ref. 9).
c
Relative energies obtained at B3LYP/6-311G**//RHF/6-31G* in kcal/mol (ref. 9).
d
Relative energies obtained with the OPLSAASEI in kcal/mol.
e
Relative energies obtained with the OPLSAA in kcal/mol (ref. 9).
f
For 1 3 X H and for 4 and 5 X CH3.
g
Optimization with a dihedral angle fixed at the torsional angle value of the corresponding HF ab initio-calculated structure.
h
Conformation shown in rotational profile of Figure 7.
i
C2C1OH angle fixed at 300.
j
C2C1OH angle fixed at 330.
k
Energy 683.3352339 au (ref. 9).
l
Energy 687.4010763 au (ref. 9).
m
Fixed dihedral angle.
n
Energy 622.3613721 au (ref. 9).
o
Energy 726.7106723 au (ref. 9).
p
RMSD for the energy difference between the RHF/6-31G* and the force field calculated relative energies of the corresponding column.

1424

Kony et al.

Vol. 23, No. 15

Journal of Computational Chemistry

Figure 5. Selected structures of the hexopyranoses under investigation. (1) D-glucopyranose, (2) D-galactopyranose, (3) D-mannopyranose, (4) D-methyl-glucopyranose, (5) D-methyl-mannopyranose.

ring oxygen atom are avoided. The shape of these two profiles
cannot be explained in terms of classical effects (i.e., electrostatic
and steric) alone. Stereoelectronic interactions favor the gauche
orientation of the O5C5C6 O6 unit.
The additional scaling factor used in the OPLSAASEI force
field significantly improves the reproduction of the ab initioderived energies of certain conformations in the investigated rotational profiles. However, there are still large errors where the
both force fields show the same discrepancies. If the improvement
in reproducing ab initio data also leads to better properties derived
in the liquid phase is tested in the following.

Figure 7. Profiles for rotation around the C1O1 bond in - (a), (b), D-glucose and - (c), - (d), D-mannose calculated ab initio at
HF/6-31G* level and with the OPLSAASEI and OPLSAA force
fields.

The performance of the OPLSAASEI and the OPLSAA force

field parameters sets in the liquid phase is evaluated in MD
simulations of -D-glucose and -D-galactose in explicit water.
Rotation around the exocyclic C5C6 bond of two hexopyranoses
is expected to lead to three conformations, namely, g, g and t,
which exist in an equilibrium (Fig. 9). The relative population of
these conformations has been evaluated using 1H-NMR spectroscopy with which different ratios are obtained for D-glucose and
3739
D-galactose.
Both intramolecular interactions and interactions

with the solvent determine the most populated structures. Therefore, reproducing the experimentally derived data will provide a
test for the newly developed force field.
For hexopyranose the relative population is directly determined
from 10-ns MD simulations. The runs differ by the initial position
of the hydroxymethylene group. The g and t starting conformations are arbitrarily chosen for the -D-glucose (Fig. 10), and the
g and t for the -D-galactose (Fig. 11). Starting from different
conformations allows assessing whether the simulation time is
long enough to obtain adequate sampling and converged values of
the evaluated properties.
The rotation around the C5C6 torsion for each run is shown as
function time in Figures 10 and 11 and the relative population
ratios are summarized together with the 3JHH coupling constants in
Table 5. Not unexpectedly, the population ratios and the 3JHH
couplings lead to the same interpretations of the simulations. The
experimental results show that -D-glucose conformers are distrib-

Figure 6. The two conformations of -D-galactose used in the torsional parameter fitting. The hydroxymethylene group present a g
(2g) and a t (2h) orientation.

Figure 8. Profiles for rotation around the C5C6 bond in -D-glucose

(a) and -D-galactose (b) calculated ab initio at HF/6-31G* level and
with the OPLSAASEI and OPLSAA force fields.

Liquid Simulation

Improved OPLSAA Force Field for Carbohydrates

Figure 9. The three minima of the hydroxymethylene groups, g, g

and t, of -D-glucose and -D-galactose, the two hexopyranose C4epimers.

uted as g:g:t 56:44:0. The ratio obtained from the OPLS

AASEI for the -D-glucose is in qualitative accordance with the
experimental results. The g and g species predominate, and
there is a negligible amount of the t conformation (69:27:4). The
simulations carried out with a g and t starting orientation of the

1425

exo-cyclic hydroxymethylene group provide the same conformational ratio (Table 5). The two simulations a and b using the
OPLSAA force field do not show the same ratio for the -Dglucose. Starting from the g conformer, the hydroxymethylene
group prefers the g orientation, while the g and t conformers
appear with equally low probability. Starting with the t conformer,
the hydroxymethylene group is distributed as g:g:t 45:51:4.
After 10ns the conformational behavior of the hydroxymethylene
group is still different in simulations a and b. Therefore, longer
simulations would be required for obtaining a converged relative
population for the OPLSAA force field. This indicates that the
conformational space is less efficiently sampled using the OPLSAA force field.
From the measured J-couplings it is known that the -Dgalactose conformers are distributed as g:g:t 17:63:20. The
OPLSAASEI results present two distributions for simulations a
and b (Fig. 11). Starting from the g conformer (simulation a), the
hydroxymethylene group is distributed as g:g:t 9:53:38.
Simulation a predicts the correct order of the conformer populations with a maximum of 15% deviating from the experimental
values. Starting with the t conformer (simulation b), the hydroxymethylene group is distributed as g:g:t 4:35:61. This distribution of populations is different from simulation a but converges
towards the behavior of a towards the end (8 10 ns, 16:50:34).
This indicates that there are still sampling problems for the OPLS
AASEI force field or that the 10ns simulation time is not long
enough for obtaining a converged population ratio.
The OPLSAA calculations present similar statistical results
for simulation a and b (Fig. 11). The predominance of the g
conformer over the t conformer is well established by this force

Figure 10. Time evolution of the dihedral angle O5C5C6 O6 of -D-glucose in dependence of the indicated starting conformation.

1426

Kony et al.

Vol. 23, No. 15

Journal of Computational Chemistry

Figure 11. Time evolution of the dihedral angle O5C5C6 O6 of -D-galactose in

dependence of the indicated starting conformation.

field. The g conformer has a too low population, which is

consistent with the OPLSAA force field overestimating the gasphase energy curves in the region of 240 360 for the galactose
(Fig. 8). Further, differences between the conformations populated

in the liquid phase by the OPLSAASEI and the OPLSAA force

field are found in the time evolution of the C1C2O2H2 dihedral angle. The sampling problems mentioned above can be recognized. The OPLSAA force field obtains about one transition

Table 5. J-Couplings and Populations of Hydroxymethylene Group Rotamers

J-coupling (Hz)

-D-glucose
H-NMR exp.
OPLSAASEIa
a:g b
b:t b
OPLSAAc
a:g b
b:t b
-D-galactose
H-NMR exp.
OPLSAASEIa
a:g b
b:t b
OPLSAAc
a:g b
b:t b
a

Relative population (%)

JH5,H6R

JH5,H6S

5.8

1.9

56

44

4.5
5.0

1.7
1.7

67
69

29
27

4
4

10.0
6.9

4.6
2.8

12
45

77
51

11
4

7.9

4.6

17

63

20

7.4
5.3

6.9
9.2

9
4

53
35

38
61

7.1
8.5

7.7
6.4

2
3

52
64

46
33

Values obtained from MD simulations using the OPLSAASEI force field.

MD starting orientation of the hydroxymethylene group.
c
Values obtained from MD simulations using the OPLSAA force field.
b

Improved OPLSAA Force Field for Carbohydrates

1427

Figure 12. Time evolution of the dihedral angle C1C2O2H2 of -D-glucose and
starting conformation of the simulation.

per nano second between different conformations, whereas the

OPLSAASEI obtains many. Considering the fact that the lifetime of a hydrogen bond formed by a hexopyranose and a water
molecule is in average between 5 and 50 ps,40 one would expect
rather more than just one rotation to occur. Visual inspection of
Figure 12 shows further that the g conformation is populated the
most by the OPLSAASEI force field whereas the OPLSAA
force field populates more g conformations.
The gas phase rotational profile for the C1C2O2H2 torsion
is shown in Figure 13. The OPLSAA force field obtains an energy

Figure 13. Rotational profiles around the C2O2 torsion in -Dglucose calculated ab initio at HF/6-31G* level and with the OPLS
AASEI and OPLSAA force fields.

for the g conformation, which is about 4 kcal/mol higher than the

ab initio calculations. This large discrepancy can be understood,
because the geometries compared in this rotational profile differ in
fragments of 1,2-ethandiol, for which already such a large discrepancy between the ab initio and the OPLSAA was found (Fig. 1).
The improvement obtained by the OPLSAASEI force field to
describe the gas phase energetics of 1,2-ethanediol is therefore
directly reflected in this rotational profile. The relative energy for
the g conformation calculated by the OPLSAASEI force field
is closer to the one calculated at the ab initio level of theory. The
large difference in the gas phase energy of the g conformations
obtained for the two force fields leads to the different populations
of the conformations observed in the liquid phase. Unfortunately,
there are no experimental data with which the most populated
conformation originating from different orientations of the C1
C2O2H2 angle could be identified.
Due to the mean field approximation used for deriving the
nonbonded interactions of the OPLSAA force field, the strength
of the intramolecular hydrogen bond formed in 1,2-ethanediol is
exaggerated, and consequently, the conformations without such a
hydrogen bond are found too high in energy (Fig. 1). These
conformations without an intramolecular hydrogen bond form in
solution more hydrogen bonds with the solvent than the conformation with an intramolecular hydrogen bond. Because, in comparison to their gas phase interaction energy, the hydrogen bonds
formed with the solvent are also exaggerated,3 the gas phase
relative energy of the conformations without an intramolecular
hydrogen bond should, therefore, be higher than the ones derived
at ab initio level of theory. In this way, a proper balance between
the intra- and the intermolecular interactions is obtained in simulations using explicit solvent. It is, however, unclear how much
higher in energy the conformations without an intramolecular

1428

Kony et al.

Vol. 23, No. 15

hydrogen bond have to be relative to the lowest energy conformations in the gas phase. This is a general problem of the parameterization concept of pair-wise additive force fields, which makes
it impossible to derive the torsional parameters systematically.
Opposite to this concept, the OPLSAASEI parameter set
does reproduce the ab initio-derived gas phase relative energy of
1,2-ethandiol conformations without an intramolecular hydrogen
bond. In this way, lower barriers for rotation are obtained for
hexopyranoses and, therefore, far more transitions between different orientations of the C1C2O2H2 angle are observed than in
the simulations conducted with the OPLSAA force field (Fig. 12).
The OPLSAASEI force field therefore describes the conformational behavior of a hexopyranose in aqueous solution in a more
realistic way than the OPLSAA force field. Because the new
scaling factors are applied within a hexopyranose only, and because Vishnyakov et al.21 found that OPLSAA describes the
conformational space of a disaccharide in aqueous solution very
well, the OPLSAASEI can be used to study polysaccharides in
solution.
The application of an additional 1,5-scaling factor makes it
necessary to also add a 1,6-scaling factor to be able to fit the
torsional parameters. Trials to fit the torsional parameters without
an additional 1,6-scaling factor and excluding the hexopyranose
conformations with an exocyclic hydroxymethylene group acting
as hydrogen bond donor or acceptor have not been successful. As
outline above, such a fitting procedure would be more compatible
with the mean-field approximation applied in pair-wise additive
force fields than applying the additional 1,6-scaling factor.

Conclusion
The OPLSAA force field for carbohydrates is modified by applying additional scaling factors for the electrostatic interactions of
the 1,5- and some special 1,6-interactions and by adjusting the
torsional parameters. This new scheme for the electrostatic intramolecular interaction has considerably improved the force field
reproducing the ab initio calculated relative energies of 1,2ethanediol and the three hexopyranoses, glucose, mannose, and
galactose. For the new force field a similar RMSD between the ab
initio calculated and the force field calculated relative energies is
obtained for the 44 carbrohydrate conformers used for the parameterization. However, it considerably reduces the largest deviation
between the ab initio results and OPLSAA force field, which are
observed when complete rotational profiles are compared. The
barriers in such profiles represent conformations with broken intramolecular hydrogen bonds and are populated in aqueous solution forming hydrogen bonds with the solvent. This indicates that
the balance between strength of intra- and intermolecular interactions is shifted increasing the strength of intermolecular interaction
energies.
In the OPLSAA force field, scaling of the 1,4-electrostatic
interactions by a factor of 2, causes an imbalance which is the
source of the largest deviation from the ab initio results. Adding
1,5-scaling factors successfully restores this balance in hexopyranoses and, by scaling the excessive electrostatic contribution to
the total intramolecular energy, improves the conformational energetics for 1,2-ethanediol dramatically. However, the addition of

Journal of Computational Chemistry

a scaling factor for 1,5-interactions makes it necessary to add a

1,6-scaling factor for particular interactions. Only with this 1,6scaling factor it is possible to restore the required balance of the
electrostatic interactions to optimize the torsional parameters for
certain hexopyranose conformations. This limits the approach
taken in this work to improve a pair-wise additive force field to the
molecular building block closed in itself, such as the one of
hexopyranoses. However, those molecular building blocks can be
combined with each other, which leads in this case to polysaccharides. If the 1,5- and the 1,6-interactions between building blocks
had to be scaled is currently being investigated by comparing the
computationally determined conformational with experimentally
derived ratios.
The solvation effect on the conformational equilibria is investigated by long time scale MD simulations on -D-glucose and
-D-galactose. This tests if the gas phase energetics is reasonable
and if the strength of the intramolecular interactions is in balance
with the strength of the intermolecular interactions. The population
ratio of the conformations originating by rotation around the
C5C6 is directly determined for -D-glucose and -D-galactose
from MD simulations. The simulations performed with the OPLS
AASEI accurately reproduce the glucose experimental results,
whereas the simulations performed with the OPLSAA force field
suffer from sampling problems. Nevertheless, results obtained for
-D-galactose indicate that there are still problems for the OPLS
AASEI force field or that the simulation time is not long enough
to obtain converged population ratios. In comparison to the OPLSAA force field, the new scheme obtains barriers for rotation of
hydroxyl groups, which are generally closer to the ones determined
at ab initio level, which leads to a higher hydroxyl transition
frequency and better solvation in aqueous solution.
The structural features of carbohydrates, and of 1,2-ethanediol
clearly demonstrate the limits of what can be handled with pairwise additive force fields. The problems encountered indicate that
a general improvement of the current molecular model can only be
achieved by replacing the mean-field approximation used to describe the polarized liquid state by an explicit treatment as done in
polarizable force fields.41 For the parameter development of polarizable force fields it will be a good benchmark to test how well
carbohydrates can be treated. Although proteins and peptides are
of larger general interest, they do not challenge the force field as
much as the structural features of carbohydrates do.

Acknowledgments
Gratitude is expressed to Dr. S. Telfer for helping to prepare the
manuscript and Dr. P. Boulet for his help in conducting the ab
initio calculations.

References
1. Lii, J.-H.; Ma, B.; Allinger, N. L. J Comput Chem 1999, 20, 1593.
2. van Gunsteren, W. F.; Berendsen, H. J. C. Angew Chem Int Ed Engl
1990, 102, 1020.
3. (a) Jorgensen, W. L. Org Chem, 1991, 4, 91; (b) Damm, W.; van
Gunsteren, W. F. J Comput Chem 2000, 21, 774.

Improved OPLSAA Force Field for Carbohydrates

4. For a brief introduction to Force Field concept, see, for example:

Connolly, M. L.; von Rague Schleyer, P.; Allinger, N. L.; Clark, T.;
Gasteiger, J.; Kollman, P. A.; Schaefer, H. F., Schreiner, P. R., Eds.;
Encyclopedia of Computational Chemistry; Wiley: New York, 1998;
vol 3, p 1698.
5. (a) Carlson, H. A.; Nguyen, T. B.; Orozco, M.; Jorgensen, W. L.
J Comput Chem 1993, 14, 1240; (b) Jorgensen, W. L. In von Rague
Schleyer, P.; Allinger, N. L.; Clark, T.; Gasteiger, J.; Kollman, P. A.;
Schaefer, H. F., Schreiner, P. R., Eds.; Encyclopedia of Computational
Chemistry; Wiley: New York, 1998, p. 1061, vol. 2; (c) Jorgensen,
W. L.; von Rague Schleyer, P.; Allinger, N. L.; Clark, T.; Gasteiger,
J.; Kollman, P. A.; Schaefer, H. F., Schreiner, P. R., Eds.; Encyclopedia of Computational Chemistry; Wiley: New York, 1998; vol 3, p
1986.
6. Ott, K.-H.; Meyer, B. J Comput Chem 1996, 17, 1068.
7. Kouwijzer, M. L. C. E.; van Eijck, B. P.; Kooijman, H.; Kroon, J. Acta
Crystallogr B 1995, 51, 209.
8. van Eijck, B. P.; Kroon, J. J Comput Chem 1999, 20, 799.
9. Damm, W.; Frontera, A.; TiradoRives, J.; Jorgensen, W. L J Comput
Chem 1997, 18, 1955.
10. (a) Glennon, T. M.; Merz, K. M. J Mol Struct (Theochem) 1997, 395,
157; (b) Glennon, T. M.; Zheng, Y.-J.; Le Grand, S. M.; Shutzberg,
B. A.; Merz, K. M., Jr. J Comput Chem 1994, 15, 1019.
11. Senderowitz, H.; Parish, C.; Still, W. C. J Am Chem Soc 1996, 118,
2078.
12. Woods, R. J.; Dwek, R. A.; Edge, C. J.; FraserReid, B. J Phys Chem
1995, 99, 3832.
13. Simmerling, C.; Fox T.; Kollman, P. A. J Am Chem Soc 1998, 120,
5771.
14. Reiling, S.; Schlenkrich, M.; Brickmann, J. J Comp Chem 1996, 17,
450.
15. Ha, S. N.; Giammona, A.; Field, M.; Brady, J Carbohydr Res 1988,
180, 207.
16. Cornell, W. D.; Cieplak, P.; Bayly, C. I.; Kollman, P. A. J Am Chem
Soc 1993, 115, 9620.
17. Jorgensen, W. L.; Maxwell, D. S.; TiradoRives, J. J Am Chem Soc
1996, 118, 11225.
18. Jorgensen, W. L.; Damm, W.; Frontera, A.; Lamb, M. L. NATO ASI
Ser C, 1996, 485, 115.
19. Jorgensen, W. L. unpublished results. Available from the parameter
file in BOSS, Version 3.5, Yale University: New Haven, CT, 1994.
20. Damm, W.; Jorgensen, W. L. unpublished results.
21. Vishnyakov, A.; Widmalm, G.; Kowaleski, J.; Laaksonen, A. J Am
Chem Soc 1999, 121, 5403.
22. (a) Daura, X.; Jaun, B.; Seebach, D.; van Gunsteren, W. F.; Mark,
A. E. J Mol Biol 1998, 280, 925; (b) Daura, X.; Gademann, K.; Jaun,
B.; Seebach, D.; van Gunsteren, W. F.; Mark, A. E. Angew Chem Int
Ed 1999, 38, 236.

1429

23. (a) Duan, Y.; Kollman, P. A. Science 1998, 282, 740; (b) Duan, Y.;
Wang, L.; Kollman, P. A. Proc Natl Acad Sci USA 1998, 95, 9897.
24. Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Gill, P. M. W.; Johnson,
B. G.; Robb, M. A.; Cheeseman, J. R.; Keith, T.; Petersson, G. A.;
Montgomery, J. A.; Raghavachari, K.; Al-Laham, M. A.; Zakrewski,
V. G.; Ortiz, J. V.; Foresman, J. B.; Ciolowski, J.; Stefanov, B. B.;
Nanayakkara, A.; Challacombe, M.; Peng, C. Y.; Ayala, P. Y.; Chen,
W.; Wong, M. W.; Andres, J. L.; Replogle, E. S.; Gomperts, R.;
Martin, R. L.; Fox, D. J.; Binkley, J. S.; Defrees, D. J.; Baker, J.;
Stewart, J. P.; HeadGordon, M.; Gonzalez, C.; Pople, J. A. Gaussian
94, Revision B.2; Gaussian Inc.: Pittsburgh, PA, 1995.
25. Csonka, G. I.; Csizmadia, I. G. Chem Phys Lett 1995, 243, 419.
26. For B3LYP performance, see, for example: Bauschlicher, C. W. Chem
Phys Lett 1995, 246, 40.
27. Cramer, C. J.; Truhlar, D. G. J Am Chem Soc 1994, 116, 3892.
28. (a) Barrows, S. E.; Storer, J. W.; Cramer, C. J.; French, A. D.; Truhlar,
D. G. J Comput Chem, 1998, 19, 1111; (b) Cramer, C. J.; Truhlar,
D. G.; French, A. D. Carbohydr Res 1997, 298, 1; (c) Barrows, S. E.;
Dulles, F. J.; Cramer, C. J.; Truhlar, D. G.; French, A. D. Carbohydr
Res 1995, 276, 219.
29. Maxwell, D. S.; TiradoRives, J.; Jorgensen, W. L. J Comput Chem
1995, 16, 984.
30. van Gunsteren, W. F.; Billeter, S. R.; Eising, A. A. Huenenberger,
P. H.; Krueger, P.; Mark, A. E.; Scott, W. R. P.; Tironi, I. G.
Biomolecular Simulation: the GROMOS96 Manual and User Guide;
Hochschulverlag AG/ETH Zurich: Zurich, 1996.
31. Maxwell, D. S.; TiradoRives, J. Potent, Yale University: New Haven,
CT.
32. Jorgensen, W. L.; Chandrasekhar, J.; Madura, J.; Impey, R. W.; Klein,
M. L. J Chem Phys 1983, 79, 926.
33. Berendsen, H. J. C.; Postma, J. P. M.; van Gunsteren, W. F.; DiNola
A.; Haak, J. R. J Chem Phys 1984, 81, 3684.
34. Haasnoot, C. A. G.; De Leeuw, F. A. A. M.; Altona, C. Tetrahedron
1980, 36, 2783.
35. Maxwell, D. S.; TiradoRives, J. Fitpar, Version 1.1.1, Yale University, New Haven, CT, 1994.
36. (a) Wladkowski, B. D.; Chenoweth, S. A.; Jones, K. E.; Brown, J. W.
J Phys Chem A 1998, 102, 5086, (b) Brown, J. W., Wladkowski, B. D.
J Am Chem Soc 1996, 118, 1190.
37. Nishida, Y.; Ohrui, H.; Meguro, H. Tetrahedron Lett 1984, 25, 1575.
38. Ohrui, H.; Nishida, Y.; Higuchi, H.; Hori, H.; Meguro, H. Can J Chem
1987, 65, 1145.
39. Nishida, Y.; Hori, H.; Ohrui, H.; Meguro, H. J Carbohydr Chem 1988,
7, 239.
40. Astley, T.; Birch, G. G.; Drew, M. G. B.; Rodger, P. M. J Phys Chem
A, 1999, 103, 5080.
41. Bernardo, D. N.; Ding, Y.; KroghJespersen, K.; Levy, R. M. J Phys
Chem 1994, 98, 4180.