Force Field Evaluation Suite
Dr. Thomas A. Halgren
Merck and Co., Inc.
P.O. Box 2000
Rahway, NJ 07065
This evaluation suite is geared to assessing the performance of molecular force
fields for (1) conformational energies and (2) intermolecularinteractions, but
the molecular structure data it includes could also be used to test the
accuracy of force-field optimized molecular geometries. The suite provides
input data and a summary of principal results for the following manuscript:
T. A. Halgren, "MMFF VII. Characterization of MMFF94, MMFF94s, and Other
Widely Available Force Fields for Conformational Energies and for
Intermolecular-Interaction Energies and Geometries," J. Comput. Chem.,
in press (expected out in April 1999).
In addition to MMFF94 and MMFF94s, the paper characterizes the CFF95, CVFF, MSI
CHARMM, CHARMM 22 (in part) AMBER*, OLPS*, MM2*, and MM3* force fields. Force
fields excluded because they were unavailable at Merck include AMBER 4, OPLS-AA,
GROMOS, MM2, MM3, and MM4. This evaluation suite has been posted so that the
community can use it to characterize these and other force fields. The data
may also be useful for developing or validating new force fields.
The manuscript makes three sets of conformational energy comparisons. The first
uses the 37 comparisons to experiment employed in the original MMFF94 paper on
this subject . It also compares the ability of theoretical methods
ranging from HF/6-31G* to GVB-LMP2/cc-pVTZ(-f) to reproduce the same
experimental data. The second set consists of 19 comparisons taken from
Gundertofte et. al  for which neither ab initio nor experimental data were
used in the development of MMFF94. The third set consists of 147 comparisons
to ab initio values obtained at the composite "MP4SDQ/TZP" level .
The comparisons for intermolecular-interaction energies and geometries employ
scaled HF/6-31G* results for the 66 small-molecule dimers used in the nonbonded
parameterization of MMFF94 . The scaling protocol is defined in a file
Input Structure Files for Conformational Energies
The following files supply input molecular structure data:
Two formats are provided: "mol2", from Tripos, and "mmd", the designation used
at Merck for BatchMin "dat" files. We chose these file formats because they
are in fairly widespread use and because they allow explicit single and
multiple bonds to be designated. Unlike file formats more commonly used at
Merck, these formats are limited in that they cannot specify formal-
charge information. However, this information is provided in another file
described below. The conf-e_37-147.mol2 and conf-e_37-147.mmd files provide
input for the first (37 membered) and third (147 membered) conformation sets.
The geometries are MP2(FULL)/6-31G* optimized. The conf-e_19.mol2 and
conf-e_19.mmd files are used for the second (19 membered) conformation set.
These files supply MMFF94-optimized geometries that should provide suitable
starting points for geometry optimization with other force fields.
Input Files Containing Conformational Energies
Reference energies are given in the following files:
The conf-e_37-147.energies file covers the first and third comparison sets.
This file gives the 5-character "conformational indices" used to label the
structure and geometry . It also specifies the total MP2/TZP energies and
the 6-31G# small-basis-set MP3 plus MP4SDQ corrections; these energies are
summed to obtain the composite "MP4SDQ/TZP" energies that were used to form
best-available ab initio conformational energy differences in the original
MMFF94 parameterization . The relationship between the 6-31G# and 6-31G*
basis sets is noted in the file. This file also contains a title-card string
for each structure that indicates its constitution and conformation.
The conf-e_37.expt, conf-e_37.mp4sdq_tzp, and conf-e_37.gvb-lmp2 files specify
the experimental, "MP4SDQ/TZP", and GVB-LMP2/cc-pVTZ(-f) conformational
energies for comparison set 1. The experimental conformational energies differ
in some cases from those used in the earlier work on the derivation of MMFF94
. An appendix to the paper, which because of space limitations has had to
be moved to the Supplementary Material (available on line from the J. Comput.
Chem. server), describes the basis for the choice of these particular
experimental values and lists some of the others that are available. The force
fields examined in the manuscript are compared to each set of reference
energies. A summary table described later shows that a given force field fits
each reference set about equally well (or poorly). This finding indicates that
all three sets provide a valid basis for assessing the accuracy of molecular
Finally, the conf-e_19.expt and conf-e_147.mp4sdq_tzp files respectively
specify the reference experimental and "MP4SDQ/TZP" conformational energies for
comparison sets 2 and 3. The experimental values for comparison set 2 were
taken from Gundertofte et al.  without further examination.
Other Data Files for Conformational Energies
As previously indicated, formal atomic charge information is not preserved in
the "mol2" input files and is represented only implicitly in the "mmd" file
through the assigned MacroModel atom types. To assist those who may wish to
utilize file formats that require explicit formal charge specifications, this
information is provided for comparison sets 1 and 3 in the conf-e_37-147.fc
file. Conformation set 2, in contrast, has no instances of non-zero formal
atomic charges. The conf_e_19.titles file lists "title card" descriptions of
the structures and geometries for comparison set 2.
Input Files for Intermolecular Interactions
The "mol2" and "mmd" input structure files provide HF/6-31G*-optimized monomer
and dimer geometries. The hbond.interactions file identifies the monomers that
form each dimer and specifies the dimer atoms that contribute to key
hydrogen-bond interactions. These specifications allow X...Z heteroatom
distances and X-H...Z hydrogen-bond angles to be computed from the input
structure files and from optimized force-field structure files derived from
them. The file also explains the procedure used to obtain the scaled QM
interaction energies and nonbonded heteroatom distances from the raw HF/6-31G*
data. The hbond.energies file lists the raw HF/6-31G* energies for the
monomers and dimers.
Other Data Files for Intermolecular Interactions
The "titles" files help to clarify the connection between the 5-character
conformational indices and the associated structures. As before, the
hbond_monomers.fc and hbond_dimers.fc files specify the atoms that carry
non-zero formal ionic charges.
These postscript files contain tables taken from the paper. Each summarizes
the overall success of the fits to experimental or ab initio data for a range
of theoretical methods.
In the conf-e_tables.ps file, the first table documents the differing abilities
of ab initio methods ranging from HF/6-31G* to GVB-LMP2/cc-pVTZ(-f) to
reproduce the experimental conformational energies of set 1. The second table
shows the ability of the force field models to fit experimental, "MP4SDQ/TZP",
and GVB-LMP2/cc-pVTZ(-f) conformational energies. The third table summarizes
the fit of the force-field conformational energies to the experimental
conformational energies of set 2, and the fourth summarizes the fit of the
force-field models to the 147 "MP4SDQ/TZP" conformational energies of set 3.
The manuscript itself also contains detailed tables that show the result given
by each theoretical method for each conformational comparison; because of space
limitations, the detailed results that the fourth table summarizes have been
relegated to the on-line Supplementary Material.
The table contained in the hbond_table.ps file summarizes the ability of the
various force fields to reproduce scaled QM interaction energies, scaled QM
X...Z heteroatom distances, and unscaled QM X-H..Z hydrogen-bond angles.
I have posted this information in the hope that it will help others to test,
or develop, additional force fields. In return, I ask those who do so to let
me know of results obtained from its use, to the extent this is feasible.
1. T. A. Halgren and R. B. Nachbar, J. Comput. Chem., 17, 587-615
2. K. Gundertofte, T. Liljefors, P.-O. Norrby, and I. Petterssen, J. Comput.
Chem, 17, 429-449 (1996).
3. T. A. Halgren, J. Comput. Chem., 17, 520-552 (1996).
4. T. A. Halgren, J. Comput. Chem., 17, 490-519 (1996).
File name Size in Bytes