AMBER Archive (2002)

Subject: Re: Potential Energy and RMS-Deviation Issues

From: Thomas Cheatham (cheatham_at_chpc.utah.edu)
Date: Thu Aug 01 2002 - 11:08:09 CDT

> I ran a 1ns production over the course of 5 200ps runs on a triplex
> (no water, PME used, Zn2+ counterions, 1fs timestep, 300K, 5kcal DNA
> and 25 kcal ion restraints throughout the run). When I plotted the
> potential energy over the course of the calculation, it stayed steady
> throughout each individual run, and dropped ~100 kcal/mol between
> each restart. As a result, my graph looks like a set of stairs. Is
> this common, or did I foul up somewhere?

Did you set IREST=1 (or in pre-AMBER7 INIT=3, IREST=1) such that new
temperatures were not assigned on restart? You probably also have to
set NTX appropriately to make sure that you are reading in the (last
step's) velocities. [Does the graph look like a set of stairs when turned
at 45 degrees, i.e. the PE dropped and then slowly ramped back up? Then
this is likely what happened; if it really looks likes stairs, i.e.
potential energy droops 100 kcal/mol and stays down on each restart, this
will require a little more tracking.]

> Also, I ran a 1ns production over 5 200ps runs on a duplex in a
> waterbox (5500 waters, PME used, Zn2+ counterions, 2fs timestep,
> 300K, no restraints). The PE was fine in this case, but the RMS-
> deviation seemed to start at 0 and drift gradually up to 1-1.5A after
> each restart. Is this supposed to happen?

If you are looking at the RMSd to the first structure in the mdcrd (from
each separate run), then the 0th frame should have an RMSd of zero.
What do you expect will happen as a function of time? As the dynamics
proceed, the structure will move away from this geometry and the RMSd
will rise. The thermal fluctuations of the DNA will lead to RMSd
fluctuations on the order of 1-1.5 angstroms (or more depending on the
length and flexibility of the system) about the average structure.

You would be better off looking at the RMSd to the initial structure and
then also calculating the RMSd to the average structure (over stable
portions of the trajectory) to get an idea about how far the structure has
moved and whether it has settled into equilibrium structure(s).

[If you are RMSd'ing to the initial structure from time 0 and the RMSd is
starting at zero each time, make sure you are using the previous restrt
file as the next sander run's inpcrd]

Good luck.

\ 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