grompp sets dihedral potentials based on order, not best match
In the force field file dihedral parameter entries like
X CT CT X
should be placed after ones with the same types that don't use wildcards. This requirement does not seem to be documented. If you place such wildcard lines first, grompp will silently choose the wildcard match and will ignore the better match. Since Amber force field (modification) files put the wildcard ones first, this leads to high risks of near-silent incorrect results when users manually convert Amber force fields.
Remove restriction on dihdraltypes order
Wildcards are allowed for dihedraltypes by grompp. But it was assumed,
with only code and no user documentation, that wildcard types come
after fully specified types. If this order was switched, grompp would
silently choose the wildcard matches. Now the match with the most real
atom type matches is chosen.
Added a paragraph to the bonded parameter section in the manual on
dihedral wildcard matching.
#7 Updated by Mark Abraham over 3 years ago
David van der Spoel wrote:
We just rediscovered this bug in old gromacs, thanks for fixing :) This should probably be added to the release notes however. It has probably led to errors in GAFF and OPLS simulations of non-proteins.
This fix is noted at http://manual.gromacs.org/documentation/5.1.3/ReleaseNotes/release-notes.html#removed-restriction-on-dihdraltypes-order. I'm open to suggestions for how to make stuff easier to find (or improvements to what is usually just my wording).
I hope to propose moving the infrastructure and source code into our repo(s) at some point, so that people can suggest their own text, and/or review at the eleventh hour before releases.
But the general problem of "how do I find whether this bug I have found (now or some time ago) was already fixed somewhere?" seems hard to solve other than choosing the right words and brute-force-thanks-to-Google.