Re: [AMBER] script generates additional dihedral terms

From: Jason Swails <>
Date: Tue, 05 Nov 2013 15:21:00 -0500

On Tue, 2013-11-05 at 10:10 -0800, Meekyum Olivia Kim wrote:
> Hi All,
> I was using Jason Swail's scripts to extract parameters from topology files, using "writeFrcmod" command, and noticed that the script gave me additional dihedral terms that are different from AMBER parameters. Other bond, angle, improper, and nonbonded terms were fine though. Would anyone be able to explain this?

There are four main reasons why torsions in an frcmod file printed by
ParmEd will not be the same as the torsions in the Amber parameter
database (or at least appear different).

First, many torsions are modeled by multiple Fourier terms (these are
indicated by a negative periodicity). The ordering of the terms in the
parameter file is not necessarily preserved in the prmtop by tleap. As a
result, the order of the torsions in a multi-term dihedral may be
different in a ParmEd frcmod file than the original force field

Second, tleap reads 4 parameters; IDIVF, PK, PHASE, and PN (see but only writes out 3
parameters. IDIVF is a dividing factor that basically indicates how
many of those torsions are present in the system. Consider ethane, for
instance, which is H3C---CH3. The H-C-C-H torsion occurs 9 times, since
each hydrogen for each carbon forms a torsion with each of the 3
hydrogens of the other carbon. So in some cases (particularly with
wild-cards), the original parameter file will have IDIVF that is not 1,
but tleap writes (PK/IDIVF) as the barrier force constant. There is no
way for ParmEd to extract the original IDIVF and PK values from the
parameter files, so it just assigns IDIVF to 1 (always) and the PK value
as the stored force constant. For torsions whose IDIVF values are not 1,
the ParmEd frcmod file will appear different than the original parameter
file, but they define equivalent parameters.

Thirdly, parameter files often contain generic torsions specified by
wild-cards. For example, see line 176 of
$AMBERHOME/dat/leap/parm/frcmod.ff12SB, which reads

X -C8-CX-X 9 1.400 0.0 3.

LEaP will match this torsion potential with any torsion whose middle 2
atoms are C8 or CX (in either order). So while the original parameter
file may only contain the above line, ParmEd will write out each torsion
term it encounters specifically since the prmtop does not indicate which
torsions were assigned by wild-cards.

Finally, the ordering of atoms in a dihedral can be reversed without
changing that dihedral at all. So while you may see


in the parameter file, this is equivalent to


The thing to watch out for the most is point 3 about the wild cards. If
you see a torsion term in the ParmEd frcmod file that you simply can't
find in the parameter database, check to see if there is a generic term
that matches with wild-cards (and keep in mind point 2 where the ParmEd
frcmod file will report PK/IDIVF instead of PK).


Jason M. Swails
Rutgers University
Postdoctoral Researcher
AMBER mailing list
Received on Tue Nov 05 2013 - 12:30:03 PST
Custom Search