AMBER Archive (2005)

Subject: Re: AMBER: Minimized structure

From: Thomas E. Cheatham, III (cheatham_at_chpc.utah.edu)
Date: Wed Jul 27 2005 - 22:47:28 CDT

> 1. How can we be assure that our structure is
> resonable after minimizations, besides looking into
> starting and ending energy and RMSD?

You cannot, depending of course on what your definition of "reasonable"
is. All you know is that you likely have moved into one of the nearest
local energy minima basins. If you look at RMSd, you will see that
suprisingly little movement in the structure is evident. Minimization
seeks to (loosely) lower the potential energy by following the gradient;
it does not search uphill and therefore will only find nearby local
minima. Despite the lack of significant movement in the structure, in
terms of energy, large changes are possible, most notably a large
reduction in the electrostatics and vdw. Things to look out for, if
starting from a seriously overlapped structure (i.e. where atoms are in
too close contact as can be checked by using the "checkoverlap" command in
ptraj), are ring fusions (imagine pulling a string through a circle such
that to pull the string out of the circle requires breaking it or serious
vdw overlap) and/or nearly infinite electrostatic energies (due to atom
overlap). In a case where the atoms get too close, electrostatics may
dominant such that with the standard potentials the minimizer will not be
able to climb up in energy to overcome a vdw overlap. [This can be
avoided by using soft core potentials or other tricks but are not standard
in the current "AMBER" force fields.]

> 2. What does GMAX means in the minimization output
> files?

This is the maximum of the gradient (force). In general you want this
value to get smaller and smaller as you minimize. There are cases where
it can get stuck, particularly with the conjugate gradient minimizer, and
cyclically increase and decrease.

> 3. Does SHAKE really affects the bondings involving H
> in minimizations?

Yes since the minimization algorithms in AMBER do not properly deal with
SHAKE constraints. It is best to turn off SHAKE in minimization to avoid
LIMIN and other minimizer failures; assuming the model has reasonable
hydrogen positions to start with, minimization will lead to a structure
that is compatible with SHAKE in molecular dynamics.

> 4. Is it ok if we use leap as a checking tool for any
> errors in structure after minimizations?

LEaP can be useful for looking at structures via edit in XLEaP, but the
program does not have any smarts or intelligence to detect errors. This
is really up to your intuition.

> 5. Any exception if we are running minimizations or MD
> on very huge structure (like 100K atoms)??

The larger the system, the less RMSd change and the more difficult it will
be to detect problems. I usually use minimization as a way to remove bad
contacts and lower the energy so MD will not blow up (due to large
forces). In a large system, problems can be masked. My recommendation is
first to look for close contacts (such as with the checkoverlap command in
ptraj) and if these are not resolved after minimization, check for odd
cases like ring fusion (where a chain gets caught in the middle of ring)
or others. If you minimize the system and then can run MD without it
blowing up, the system likely does not have serious overlap problems or
major instabilities.

With experience, you will start to learn what energies are reasonable
(i.e. bond/angle/dihedral energies that are positive but not huge,
electrostatic energies that are likely significantly negative,
particularly if water is present, vdw that are not too big, etc). If you
want to learn more, the AMBER tutorials are an excellent place to start
since you can reproduce what others have seen. Note that what energies
one observes will depend on what force field is applied so ideas learned
about what to expect from one force field (in terms of energies) may not
transfer to another depending on the similarity of the underlying
potential.

Good luck.

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu