AMBER Archive (2003)

Subject: Re: A problem in Sander production dynamics

From: Thomas Cheatham (cheatham_at_chpc.utah.edu)
Date: Mon Apr 14 2003 - 16:24:31 CDT


> However, when I've run the production dynamics, the RMSd
> result between the trajectories and the reference structure
> (from PDB structure solved by NMR) shows a strange periodical
> repeat every 120ps, and the repeat curves are almost the same.
...
> The following is my input for production dynamics, and a rmsd figure
> is shown here http://bioinfo.ndhu.edu.tw/~yuann/TMP/sander7.gif

As Carlos Simmerling mentioned already, > 16 angstroms RMSd in a few ps is
a rather large number (representing a significant change in the
structure). This could be indicative of an imaging artifact, namely that
the drug or one of the DNA strands is crossing the (periodic) box and is
being imaged back since IWRAP=1.

[As an aside, looking at your input file:

> ntc=2, tol=0.0000001, ntf=1,

ntf should equal ntc (NTF=2, NTC=2) i.e. you shouldn't include the bond
from hydrogen forces if SHAKE on hydrogens is turned on.
]

However, beyond the potential imaging issue, your RMSd plot displayed via
the above link is *very* surprising. It is incredibly periodic (but not
exactly so in terms of the magnitudes, but perfectly in terms of the
times). This is unusual. Moreover, there are very few oscillations
evident in the short time scale (1-10 ps) related to expected thermal
motion (~0.5-1A range) if you look at the minor bumps between the peaks
and valleys. In spite of the small short time-scale oscillations, there
are larger repeating oscillations evident on the longer time scale. The
lack of short time scale oscillations reminds me of the "block of ice"
problems as if the molecule is simply translating through the periodic
box or repeatedly sampling particular modes of motion. If you were having
imaging issues, I would expect large instantaneous jumps in the RMSd from
one step to the next not what is shown in that plot.

Otherwise, if it is not a motional artifact (in spite of the NCSM
variable being set), I cannot explain it at all.

How long did the MPI jobs runs (i.e. in terms of ps); multiples of 120 ps?
Is this system in vacuo?

I think Prof. Simmerling's suggestion of looking at a movie (prior to
any RMS fitting, imaging or any modification to the coordinates i.e. on
the raw trajectories produced by sander) is worthwhile.

[Another possibility is that the prmtop you are analyzing with does not
correspond to the trajectory; make sure that the prmtop matches what you
started with and you haven't done any stripping of the trajectory prior;
this could also lead to a pseudo-periodicity artifact since the trajectory
may come back in sync every 120ps worth of frames.]

Let us know if we can be of more help.

--tom

\ Thomas E. Cheatham, III (Assistant Professor) College of Pharmacy
| Departments of Medicinal Chemistry and of University of Utah
| Pharmaceutics and Pharmaceutical Chemistry 30 South 2000 East, Room 201
| & Center for High Performance Computing Salt Lake City, Utah 84112
|
| e-mail: tec3_at_utah.edu phone: (801) 587-9652 FAX: (801) 585-9119
\ http://www.chpc.utah.edu/~cheatham Offices: BPRP295A / INSCC 418