AMBER Archive (2003)

Subject: RE: How to do a rigid solvent model?

From: Yong Duan (yduan_at_udel.edu)
Date: Sun Jul 06 2003 - 14:59:23 CDT


Carlos is definitely right ...
I was hoping his solvent molecules are not that big and they are
single-residue molecules ...

yong

-----Original Message-----
From: Carlos Simmerling [mailto:carlos_at_csb.sunysb.edu]
Sent: Saturday, July 05, 2003 4:07 PM
To: Yong Duan; 'Marco Aurelio Correia Preto'
Cc: amber_at_heimdal.compchem.ucsf.edu
Subject: Re: How to do a rigid solvent model?

CAUTION-
the "faking the H name" approach may work sometimes, but
it can be very dangerous. My understanding of the MPI shake
code is that all atoms in a bond to be SHAKEn must
belong to the same residue. If the bonds involve hydrogen,
this works fine. If not, it doesn't try to figure out if the bonds
are coupled or not. For ntc=3 (all bonds) it clearly will
be incorrect The parallelism for SHAKE is such that the
iterative procedure of moving atoms will be a problem if
different processors are moving the _same_ atom to
fix the bond lengths for different bonds. The code for
this can be changed (we once modified it in our local version
of AMBER6 to use ntc=3 in parallel) but it is not that
way by default.

In other words, for a solvent molecule that has all of
the atoms in 1 residue, you should be fine in faking
the bonds by putting them all in the "bonds to hydrogen" list.

Otherwise, you should make very sure that you understand
that part of the code if you try to circumvent the limitation
of MPI sander only allowing ntc=1 or 2. It is there for good
reason!

feel free to correct me if I am wrong...

Carlos
===================================================================
Carlos L. Simmerling, Ph.D.
Assistant Professor Phone: (631) 632-1336
Center for Structural Biology Fax: (631) 632-1555
Stony Brook University Web: http://comp.chem.sunysb.edu/carlos
Stony Brook, NY 11794-5115 E-mail: carlos.simmerling_at_stonybrook.edu
===================================================================

----- Original Message -----
From: "Yong Duan" <yduan_at_udel.edu>
To: "'Marco Aurelio Correia Preto'" <mcpreto_at_fc.up.pt>
Cc: <amber_at_heimdal.compchem.ucsf.edu>
Sent: Saturday, July 05, 2003 2:48 PM
Subject: RE: How to do a rigid solvent model?

>
> Just to follow-up what Carlos was saying,
>
> In AMBER, SHAKE is parallelized for H- atoms only. So, once you invoke
> SHAKE you can constrain bonds with H atoms on parallel platforms. This
> rule does not apply to single-CPU platforms. In the later case, you
can
> use SHAKE to constrain all bonds (ITC=3?).
>
> Assume you use parallel platforms, if only the bonds connecting to H
> atoms are rigid in you solvent model, you are ok. Just set ITC=2. In
> case, other bonds are also required to be rigid, you can "fake" the H
> atoms. I tried couple of times, SANDER did not really know the
> difference. Of course, this can be confusing to a third party who
> happens to look at your PDBs and you may need to write a script to
> convert the atom names back. In the end, we probably should make it
more
> flexible to allow users to specify a list of atoms whose bonds are to
be
> constrained.
>
> yong
> -----Original Message-----
> From: Carlos Simmerling [mailto:carlos_at_csb.sunysb.edu]
> Sent: Friday, July 04, 2003 1:25 PM
> To: Marco Aurelio Correia Preto
> Cc: amber_at_heimdal.compchem.ucsf.edu
> Subject: Re: How to do a rigid solvent model?
>
>
> Large force constants are not a good idea-
> they increase the frquency of the motion and
> thus require smaller time steps for stable
> MD simulation. The TIP3P water used in AMBER
> is kept rigid by SHAKE, you might want to try
> something similar.
>
>
> > On Fri, Jul 04, 2003, Marco Aurelio Correia Preto wrote:
> > >
> > > We are doing some molecular dynamic runs of small non-standart
> peptides
> (in
> > > water) and we would like to do the smae dynamics in other
solvents!
> There
> > > are a couple of potential for organic solvents, but some of them
are
> rigid
> > > one! How can we do this in Amber (version 6 is the one we have)?
Can
> we
> use
> > > very large force constants to force a rigid conformation?
> > >
> >
>
>
>