Gibbs, PME, DynModW, ISDX0, etc.

From: Michael Jakusch <mjakusch_at_pharma.anbi.ethz.ch>
Date: Thu 20 Sep 2001 14:39:47 +0200

Dear Amber users/developers!

I am preparing some work with gibbs; however there are some issues I
could not clarify from the manuals and literature:

1. Particle Mesh Ewald:

 The manual says:
 
   "As yet, there is not much experience with the PME in Gibbs, so it
   should be considered an experimental option. ... It has been tested
   only with "standard" window growth"
 
 Did this situation change in the meantime? I would definitely like to
 use TI+PME+DynModW!


2. Treatment of vanishing atoms:

 - Is there a recommended value or range for IDSX0 (factor for the
   mixed VdW radii)?

 - There seems to have been some controversy about the shrinking of
   bonds to dummy atoms in order to improve sampling. Is it now
   generally recommended to shrink those bonds or to leave them as they
   are?

 - In the AMBER FAQ I found that - if bond shrinking is employed at
   all - the length of the bond should not be decreased below ~0.2 Ċ
   for stability reasons. This is explained in terms of an increased
   frequency of the corresponding valence vibration. I would like to
   understand, why this frequency changes at all when only the
   equilibrium bond length is changed but the force constant is not.


3. Control of sampling and convergence:

 - I thought that in order to estimate if the simulation hat already
   sufficiently sampled phase space at a certain value of l
   ("l"="\lambda") it might be a good idea to monitor the "running
   values" of dV/dl and <dV/dl> (for TI). However, there are only the
   final values of <dV/dl> for each l printed in the POUT file. Did I
   not find the necessary input flag, is this not implemented, or am I
   somehow conceptually wrong here?

 - Is it possible to write a restart file at the end of each \lambda
   interval, so if one decides to simulate for some more time one can
   start again there and does not loose the previously accumulated
   values of <dV/dl> ?

 - I suppose that the default control parameters for the DynModWin
   algorithm are safe to start with. If not, what parameter should be
   given a differnt value?

 - Is there any experience concerning the necessary steps for
   equilibration and data collection for a "minor" mutation (GLN->ASN)
   in a protein (using PME+DynModWin)?


Thanks in advance for your advice!

Michael



-- 
Dr. Michael Jakusch
ETH - Swiss Federal Institute of Technology
Departement of Applied Biosciences
Winterthurerstrasse 190
CH-8057 Zürich
Switzerland
Phone: +41.1.635 60 71
Fax:   +41.1.635 68 84
email: michael.jakusch_at_pharma.anbi.ethz.ch
Received on Thu Sep 20 2001 - 05:39:47 PDT
Custom Search