AMBER Archive (2009)

Subject: Re: [AMBER] Disappearing common atoms in Amber 10 with TI

From: steinbrt_at_rci.rutgers.edu
Date: Mon Jul 20 2009 - 07:05:24 CDT

Hi Ilja,

> What is the reason that makes removing X-H bond constraints so desirable
> for TI calculations?

SHAKE bond constraints change the atom positions. If an atom is SHAKED on
one process and not SHAKED on the other, it would loose coordinate
synchronisation. One could rewrite the SHAKE algorithm to check for such
things, but it was easier to just remove the SHAKE constraint on those few
(usually zero or just one) X-H bonds that cross between common and SC
atoms.

> Also, I would like to use soft cores, but I was under the impression that
> you do not have any of the soft Coulomb potentials implemented. This makes
> the vdW soft core potential less useful for me. Please let me know if you
> actually have the soft Coulomb implemented but undocumented, and how to

They are implemented in the developers version of Amber11 if you have
access to that (if not, you could ask Dave Case for an advance copy of
sander11...). With the new version, you can do arbitrary one-step
transformations, but it is not well tested yet...

But even in Amber10 SC potentials are useful. You have to do your
transformation in several steps anyway:

1) remove charges from disappearing atoms, leaving their vdw core intact
2) remove and appear any changing atoms
3) add charges to any freshly appeared atoms

Sum up 1-3 to get the total deltaG

step 2 not only benefits from soft core potentials in terms of convergence
and ease of setup, it also allows appearing and disappearing atoms
simultaneously, which you cant do with the old klambda>1 approach.

Kind Regards,

Thomas

Dr. Thomas Steinbrecher
BioMaps Institute
Rutgers University
610 Taylor Rd.
Piscataway, NJ 08854

_______________________________________________
AMBER mailing list
AMBER_at_ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber