Re: amber-developers: Question concerning wildcards in dihedral definitions.

From: Thomas E. Cheatham, III <>
Date: Wed, 29 Jun 2005 15:10:53 -0700

> I think(?) that all of the stuff Ross picked up on is unique to
> that is the only place where there are the specific sugar-based torsions
> (added by Junmei) that potentially conflict with wild-card torsions. As
> I understand it, there is no corresponding conflict in parm94.dat.

I will verify that for parm94.dat the dihedral terms are identical with
LEaP before and after the "copy" unit trick of Ross with nuc.pdb except in

exactly the opposite order, i.e. 1-2-3-4 vs 4-3-2-1 for ALL torsions with
hydrogen and some of the torsions w/out hydrogen. The impropers are in
the same order. This order change makes comparing things a distinct pain.

As far as PLEP goes, everything is equivalent to LEaP EXCEPT for change of

one improper:

XXX I 32: 1.100 3.14 2.0 :1.N7 :1.N9 :1.C8 :1.H8
XXX B 47: 1.100 3.14 2.0 :1.N9 :1.N7 :1.C8 :1.H8

which alters the atom order. This does alter the energy by a small amount

(due entirely to this torsion) but should not be disasterous for


Searching through the archives/WWW page I cannot find discussion of
flipping the order on this improper.

In contrast, with parm99 as Ross points out, we see differences before
and after the LEaP "copy" unit:

LEaP pre-copy DIHEDRAL ENERGY = 17.3691
LEaP post-copy DIHEDRAL ENERGY = 17.6827

The pre-copy and PLEP differ *only* by an improper switch exactly as
was seen with parm94 i.e. N9 N7 C8 H8 swapped with N7 N9 C8 H8.

Post-copy, the differences are more significant with additional torsions
added, specifically H1-CT-CT-OS (or OS-CT-CT-H1) with a 3-fold 0.156 peak
value. So, "pre-copy" (what most people do when building) is more correct

in terms of agreement with PLEP and searching through a bunch of my LEaP
created parm99 prmtop's I only see the V1 term and not the V3 (as seen

What I guess is happening is that we are matching X-CT-CT-X and
H1-CT-CT-OS post the unit copy in LEaP but *not* prior to the copy.
(however read on)

[Replying back to DAC, despite the fact that frcmod requires a strict
match, given the old style AMBER where a single torsion could have
multiple terms, if it was a specific match, it overrode everything, or at
least that is what PLEP does. However, just because it WAS that way,
doesn't mean it was this way when Junmei did parm99 since he was likely
using LEaP by this point. Looking at parm99, a number of the torsions do
not make sense to me UNLESS both are applied, i.e.

F -CT-CT-F 1 1.20 180.0 1. Junmei et al,
H1-CT-CT-OS 1 0.25 0.0 1. Junmei et al,
H1-CT-CT-F 1 0.19 0.0 1. Junmei et al,

i.e. the F-CT-CT-F alone as above just favors 0 (cis) where we expect a
gauche-anomeric effect favoring g+ or g-. If you include the X-CT-CT-X w/

a peak of 0.56 (i.e. 1.4/9), then you get a potential that nearly flat
from g- to g+ and disfavors trans which makes sense (since the 1-4
interaction will disfavor cis). With the H1 you do not expect a
gauche-anomeric effect; the v1 potential alone will favor trans by
a small amount (which is OK but not obvious) whereas having the V1 and V3
puts a potential that has a minima at g+, g- and trans with trans

So, I agree that it makes sense to apply both torsions, however this is
NOT done by LEaP (pre-copy of a unit) or PLEP so I still do not understand

what is correct.

With the new parm99 of DAC, the results are equivalent before and after
the copy unit in LEaP, but now the energy does not agree with parm99.dat
(either before or after the copy)


as nine additional V3 torsions are added to the pre-copy original parm99
results; OS-CT-CT-H1, OS-CT-CT-HC, H1-CT-CT-OH, i.e. torsions on H4', H3T,

H2'? in addition to H3', H5'? we saw previously. This confuses me even
more as why after the copy did LEaP not add the V3 to ALL the X-CT-CT-X
interactions involving H i.e. only to H1-CT-CT-OS and NOT HC-CT-CT-OS or

Let me know if anyone needs clarification. (I certainly do?)

Received on Wed Apr 05 2006 - 23:49:55 PDT
Custom Search