Re: [AMBER-Developers] Leap inconsistencies with wildcard torsion (surprise)

From: Jason Swails <>
Date: Fri, 13 Nov 2015 15:16:42 -0500

On Fri, Nov 13, 2015 at 12:42 PM, Niel Henriksen <> wrote:

> Hi All,
> Thanks for your analysis. I am actually less worried at the moment about
> whether the specific torsion SHOULD overwrite the generic torsion or not.
> (Although it would be nice to have this cleared up.) What worries me is
> that the behavior is not consistent within the same molecule for identical
> chemical groups. Some of the "hc-c3-c -o" connectivities show that the
> generic torsion IS getting removed, whereas in other "hc-c3-c -o"
> connectivities it is NOT removed ... all within the same molecule.

​A minor nit here -- I don't think terms ever get *removed* -- it's just
that the generic terms are not always added first, it would appear. Only
if the generic terms are added first, and the specific ones added later,
would the periodicity 2 values crop up.

​As this is, if not explicitly documented, *known* behavior, we really have
to fix gaff.dat to specifically override generic torsions​

> ​​
> ​​
> Perhaps there is some subtle reason why this behavior is correct. But if
> ​​
> it is a bug, then it doesn't matter what we decide is the correct behavior
> ​​
> because it appears that LEaP is already not following the "rules"
> ​​
> consistently.

​Actually, according to tleap there are no rules here -- we are firmly in
the realm of undefined behavior. The rule is imposed on the construction
of the parameter files -- not on tleap. If a torsion type has a matching
generic term, then any specific term that matches a subset of the generic
term's applicability has to override every periodicity term that generic
term defines *in addition to* whatever other terms it wants to add.

So the problem is really that gaff.dat does not contain a 3rd line for
c3-c3- c- o that defines a term with periodicity 2 and force constant 0.
In that case, every instance in the prmtop will have the same 3 terms.

I will see if I can work up a quick script to highlight offending specific
terms in the gaff database and get this file fixed.

All the best,

Jason M. Swails
Rutgers University
Postdoctoral Researcher
AMBER-Developers mailing list
Received on Fri Nov 13 2015 - 12:30:05 PST
Custom Search