[MMTK] Re: MInimization in MMTK using Amber

Konrad Hinsen hinsen@cnrs-orleans.fr
04 Apr 2002 11:52:30 +0200


"Andrei A. Korostelev" <korostel@chemmail.chem.fsu.edu> writes:

> Discover makes use of Amber as MMTK does. I hope there is no
> difference in minimization algorithms between Discover and MMTK. As a

There sure is. Although there are only a few commonly used algorithms
for minimization, each of them actually refers to a broad category
of similar but not quite identical versions.

> Now I apply the same test to a little more complicated molecule. The
> difference between programs is significant now. So I wonder why there
> is a difference between two programs using the same force field?

Because they converge to different minima. In your simple test case
there was probably only one minimum near tjhe initial structure, for
a realistic case there are many.

You could try starting with steepest descent. Although not efficient
close to a local minimum, this algorithm has the advantage of always
moving towards the minimumn that is closest to tne initial structure.
Conjugent gradients comes with no such guarantee.

> Here is the set of more concrete questions:
> 1) Are there any distance cutoffs, switches (for electrostatics?) etc.
> in MMTK that would influence minimization?

Of course, all force field parameters. But you should be able to
define the same set for both MMTK and Discover.

> 2) Is it right that if I don't specify algorithm for electrostatics
> (Ewald etc.), MMTK uses Coulombic potential with constant (not
> distance-dependent) dielectric as a default?

Yes.

> 3) How can I set and check a value for dielectric constant?

There is no provision for that in standard MMTK, but it is simple
to add, I should have an example somewhere...

> 4) How can I check partial charges ascribed to each atom?

print atom.charge()

> 5) Would you recommend using MMTK2.2b instead of MMTK2.0 I'm using

Yes, definitely. MMTK 2.0 is very old by now, nd MMTK 2.2 would long
have been released as "official" if there hadn't been some
problems with the new installation procedure. These are now solved,
all that is missing for an official release is the time to package
and check everything.

> now? Where can I download 2.2b version (link to
> ftp://dirac.cnrs-orleans.fr/pub/MMTK/MMTK-2.2b1.tar.gz is not
> functional)?

That's

  ftp://dirac.cnrs-orleans.fr/pub/MMTK/MMTK-2.2b10.tar.gz

and it has been used by several people already, so it should work.

Konrad.
-- 
-------------------------------------------------------------------------------
Konrad Hinsen                            | E-Mail: hinsen@cnrs-orleans.fr
Centre de Biophysique Moleculaire (CNRS) | Tel.: +33-2.38.25.56.24
Rue Charles Sadron                       | Fax:  +33-2.38.63.15.17
45071 Orleans Cedex 2                    | Deutsch/Esperanto/English/
France                                   | Nederlands/Francais
-------------------------------------------------------------------------------