ORGANIC CHEMISTRY
DIVISION
COMMISSION ON PHYSICAL ORGANIC CHEMISTRY
Guidelines for Publication of Research Results
from Force-Field Calculations
(Technical report)
Douglas J. Raber
and Wayne C. Guida
Membership of the Working Party during
the preparation of this report
Preamble: For the publication of research results, the chemical sciences
community has had a long history of requiring authors to provide sufficient
data so that their research results and procedures can be (1) understood,
(2) critically evaluated, and (3) replicated by other competent scientists.
The emergence of computational chemistry as a distinct area of research presents
new challenges in defining criteria to meet these obligations. While much
of the long-standing paradigm for experimental chemistry can be directly transferred
to computational chemistry, some differences are apparent. A computational
study does not give a product for which one can measure physical properties,
nor are percent yields and recoveries available to demonstrate experimental
success. Nonetheless, it is imperative that computational results be able
to withstand the same scientific scrutiny as experimental ones.
Like all fields of scientific endeavor, computational
chemistry is also a dynamic science. The continuous and dramatic improvements
in computational algorithms and increases in computing power over the
last decade have made possible the study of chemical problems for which
solutions by computational means previously were unattainable. Moreover,
advances in computer technology have also changed the way these computational
studies are carried out. For any new study, the traditional search for
the nearest energy minimum may no longer be adequate, fewer assumptions
and approximations may be acceptable, and even the nature of the data
to be stored and reported may have evolved. For example, many computer
algorithms have become sufficiently fast and convenient that it is more
efficient to repeat some part of the overall calculation than to save
and record the corresponding data that it generates.
This document has been developed to provide guidance to
chemists who employ computations of molecular structure, properties,
reactivity, and dynamics as either a part or as the main thrust of a
research report. It is derived in part from earlier work carried out
by the Provisional Section Committee on Medicinal Chemistry of IUPAC
(Gund, P.; Barry, D.C.; Blaney, J.M.; Cohen, C.N. J. Med. Chem., 1988,
31, 2230-2234).
Guidelines
1. If a paper mentions any calculations that are integral
to the study (i.e., the precise quantitative results are central to
the conclusions drawn) computational details must be described.
2. Reproducibility of computations should be a defining
goal in reporting computational results. Sufficient technical details
should be provided to allow a reader to reproduce the results of any
calculations. Note that strict reproducibility of computations may be
precluded by differences among computers (e.g., differences in operating
systems, host-based algorithms, machine precision, etc.), software (e.g.,
installation procedures and programming modifications), and applications
(e.g., for probabilistic methods such as Monte Carlo calculations).
3. A description of the results of any published calculation
must be available for scientific scrutiny by other investigators. Ordinarily
this will best be achieved if the authors place full details of the
calculation (e.g., structures, energies, frequencies, dynamical quantities,
etc.) in a repository that is accessible via the internet. Otherwise,
it is the responsibility of the investigator to make such information
available upon request. Depending on the specific circumstances, the
appropriate data to be deposited might be final output structure(s)
or the input structure(s) that would allow another investigator to replicate
the computational experiment.
4. Electronically deposited structures should be written
in a format that is machine readable by readily available modeling or
translating programs.
5. An adequate specification of the computer program used
to carry out the calculations is essential. If the calculations were
carried out with commercially available software, the investigator must
list the source, version number, and force field employed. Any program
modifications and any differences from the published version (including
changes in parameters) must be fully described or available to other
investigators upon request. Changes in parameters, program options (e.g.,
dielectric constant, nonbonded cutoffs, etc.) affecting reproducibility
of the calculation, and program constraints must be fully described
in the article. If the program is not available, commercially or otherwise,
the authors must specify sufficient detail that the calculation could
be reproduced.
6. The issue of convergence must be adequately addressed.
An iterative calculation is considered to have converged when further
iterations will not significantly alter the results. Convergence criteria
should be reported (e.g., energy change per iteration or energy gradient
for geometry optimization of a single structure).
Membership of the Working Party during the preparation
of this report: Wayne C. Guida (Novartis Pharmaceuticals Corporation)
and Douglas J. Raber (National Research Council, USA), Co-Chairs. Members:
James Crabbe (University of Reading), Jan Hermans (University of North Carolina),
Joseph B. Lambert (Northwestern University), Kenny B. Lipkowitz (Indiana-Purdue
University), George W. A. Milne (National Institutes of Health), Philip S.
Portoghese (University of Minnesota), W. G. Richards (University of Oxford),
Donald G. Truhlar (University of Minnesota), George Schatz (Georgia Institute
of Technology), Paul von R. Schleyer (University of Georgia), Andrew Streitwieser
(University of California, Berkeley), John Zdysiewicz (Australian Journal
of Chemistry)