Re: AMBER: Questions on Phi, Psi angles produced by pmemd/sander7

From: Robert Duke <>
Date: Sat, 23 Aug 2003 08:41:49 -0400

Chen -
1) You probably saw the note on the reflector about the COM velocity removal
bug in pmemd. I presume you are doing COM velocity removal? If so, there
could be a slight retardant affect on atom 1 associated with this bug, and
that is most probably an atom associated with the N terminus of your
peptide. I think it would be interesting to check out if applying the
bugfix makes a difference.

2) Also, there is a bug in sander7 mpi minimization, (requires that you have
shake enabled, which I believe is not recommended, but I have seen it done
in tutorials), (bugfix 42). This could have an effect on the sander 7
starting point (making it different in the wrong direction).

3) Are you using sander 7 compatibility mode in pmemd? If not, there may be
sander 6/7 differences; step integration is done differently, for one thing.
Also, you need to look carefully at what mdin options are actually getting
used in the runs. The defaults in sander 6 and 7 are different, and pmemd
attempts to track the defaults in 6 vs 7 mode. This means though, that 6
results may well be different than 7 results.

4) The molecular and atomic virials are handled differently in sander 6 vs
sander 7; this really only shows up if you introduce bonds between molecules
(like a disulfide intermolecular link). PMEMD uses the sander 6 convention
in both sander 6 and sander 7 compatibility mode because neither is
particularly more correct. I really doubt this is directly involved, but
will have a slight impact on the evolution of trajectories.

5) The dihedral code has had some efficiency modifications, but is
essentially identical to the sander 6 code, and produces identical energy
outputs for the short timescale one would expect. I'll look at deltas
again; it is also possible some slight changes were made here in sander 7.
I would expect any bugs in this stuff to show up fast, though.

6) Others can better address variability in Ramachandran plots derived from
different samplings of phase space. What happens if you continue the runs
and sample later chunks of trajectory? What happens if you split the
existing runs? Someone please comment on this.

Thats all that immediately comes to mind. Feel free to send me any relevant
data about run conditions.

Regards - Bob

----- Original Message -----
From: "Chen, Ya" <>
To: <>
Sent: Saturday, August 23, 2003 12:47 AM
Subject: AMBER: Questions on Phi, Psi angles produced by pmemd/sander7

> Hi, everyone,
> I did a short run (<1ns) for a system of a small peptide in TIP3P water
box using both pmemd and sander7 and found that the restart structures from
equilibration (minimization followed by 75ps md) by pmemd and sander7 were
quite different. The difference was maintained throughout production runs.
Pmemd did not disrupt the molecule as much as sander7 when RMSDs are
considered. But when examining individual residues, I found that for some
residues, (Gly-Arg, in a beta-turn,) pmemd set disallowed (in Ramachandran
plot) phi-psi values while phi-psi’s from sander7 are more allowed. By the
end of the simulation, those disallowed phi-psi’s were not changed yet.
> My questions:
> Is it normal that some torsion angles are disallowed throughout
> What is the possible cause of the inconsistency between pmemd and sander7
for my system?
> The system was created with ff99 (Amber7) and temp=300 in production runs.
The input files are identical. The starting structure (before minimization)
was resolved from NMR spectra, without disallowed torsion angles.
> Thanks for your help.
> Chen
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to
> To unsubscribe, send "unsubscribe amber" to

The AMBER Mail Reflector
To post, send mail to
To unsubscribe, send "unsubscribe amber" to
Received on Sat Aug 23 2003 - 13:53:00 PDT
Custom Search