The document discusses the development of a third-generation hydrogen-bond and three-body dispersion correction for the semi-empirical method PM6 (PM6-D3H+) in GAMESS, aimed at improving predictions for biomolecular systems by considering non-covalent interactions. It details the implementation of dispersion and hydrogen-bond corrections and compares the performance of various models through RMSD and MD values. Additionally, it introduces MolCalc, a web application for estimating molecular properties, designed for teaching purposes, which allows students to build molecules and gain insights into chemical structures and properties.

1. Jimmy Charnley Kromann
University of Copenhagen, Department of Chemistry
jimmy@charnley.dk, computeranchemistry.blogspot.com
Third-Generation Hydrogen-Bond and Three-Body Disper-
sion Correction for Semi-Empirical Method PM6 (PM6-
D3H+) in GAMESS
The ability to produce accurate quantum mechanical (QM) predictions for large-scale bio-
molecular systems has the potential to bring new insight into several scientific fields. Using
semi-empirical method PM6 we can get fast predictions. However PM6 is parameterized
for primarily covalent molecules, with little dispersion and for bio-chemical systems non-
covalent interactions is the key for getting good results, which is why the energy model
needs dispersion and hydrogen-bond correction.
Similar method already exist (PM6-DH+) which is implemented in MOPAC[2], with
the COSMO solvation method. This implementation will be in the software package
GAMESS[1], which gives improvement in geometry minimizations, multi-threading, and
PCM solvation method.
Dispersion
We chose to use Stefan Grimme’s three-body dispersion correction (dftd3)[4] because alter-
native dispersion corrections uses strong atom-type parameters and because this program
was already implemented in GAMESS. Parameters was found for PM6 in PM6-D3H[5].
Hydrogen-Bond
We chose to use Martin Korth’s Hydrogen-Bond correction (-DH+)[3] because of the con-
tinuous derivatives.
Preliminary Results
We re-parameterized the hydrogen correction to fit the new dispersion correction, which
resulted in very little deviation. Hydrogen-Bond Parameters were found using scan of
parameters and selected with the lowest RMSD values for the S22 + S66 hydrogen-bond
subsets from BEGDB
The resulting Root-mean-square deviation (RMSD) and Mean deviation (MD) in respect to
the CCSD(T)/CBS interaction energies from the S22 and S66 sets is shown here in kcal/mol.
MPM6 GPM6 MPM6-DH2 MPM6-DH+ GPM6-D3H+
Full S22 + S66
RMSD 3.36 3.36 0.83 0.78 0.76
MD 2.87 2.87 0.31 0.13 0.06
Hydrogen subset
RMSD 4.52 4.52 1.07 0.88 0.93
MD 3.95 3.96 0.37 -0.11 -0.13
Dispersion subset
RMSD 2.99 2.99 0.45 0.45 0.44
MD 2.67 2.67 0.05 0.05 0.05
Minimization of conformations of Cucurbit[7]uril (CB7) using PM6-DH+/COSMO in
MOPAC and PM6-D3H+ in GAMESS, showing the gradient norm and corresponding
amount of imaginary vibrational frequencies.
PM6-DH+/COSMO PM6-D3H+/PCM
conf grad norm i vib grad norm i vib
1 0.900 21 0.0000897 0
2 0.820 4 0.0001028 1
3 0.987 6 0.0000995 0
4 0.928 7 0.0000739 1
References
[1] GAMESS, www.msg.ameslab.gov/gamess
[2] MOPAC, www.openmopac.net
[3] M. Korth J. Chem. Theory Comput. 2010 12 33803-3816
[4] S. Grimme and J. Antony and S. Ehrlich and H. Krieg J. Chem. Phys. 2010, 132, 154-104
[5] S. Grimme Chem. Eur. J. 2012, 18, 9955-9964
Acknowledgments
Jan H. Jensen (University of Copenhagen), Anders S. Christensen (University of Copen-
hagen), Martin Korth (University of Ulm) and Casper Steinmann (University of Southern
Denmark)
Molecule Calculator (MolCalc)
A Web Application for Fast Quantum Mechanics-Based Estimation of
Molecular Properties
dgu.ki.ku.dk/molcalc
The Molecule Calculator (MolCalc) is a web application that allows students to build small
molecules and estimate properties such as molecular structure, thermodynamic properties,
vibrational frequencies, vibrational modes, molecular orbitals, orbital energies and solva-
tion energies, in a matter of seconds or minutes, depending on the size of the molecule. A
video demonstrating these features can be found on the Web site.
MolCalc was designed to be used for teaching and for assignments in which the students
build their own molecules and estimate the molecular properties, as opposed to reading
from tables in textbooks. MolCalc was designed to run fast, and therefore, the estimated
molecular properties will not match experimental values exactly and, in some cases, will
be quite different. The idea was to have students develop a chemical intuition about how
molecular structures affects molecular properties, without performing the underlying cal-
culations by hand (which would be nearly impossible for all but the simplest chemical
systems).
Similar to a pocket calculator or a symbolic math program (such as Mathematica or
MAPLE), MolCalc allows an instructor to assign higher-level chemical problems that are
not practically possible to solve otherwise.
Code
MolCalc is OpenSource and is distributed through GitHub under the GPL license
github.com/jensengroup/molcalc. You must obtain a copy of the GAMESS code sepa-
rately.
References
Jan H. Jensen and Jimmy C. Kromann, The Molecule Calculator: A Web Application for
Fast Quantum Mechanics-Based Estimation of Molecular Properties, Journal of Chemical
Education, 2013, DOI: 10.1021/ed400164n
Acknowledgments
Maher Channir (University of Copenhagen) and Jan H. Jensen (University of Copenhagen).
The development of MolCalc is supported by the University of Copenhagen through the
Education at its Best initiative (Den Gode Uddannelse).