[MMTK] Difference in number of Torsion parameters between MMTK and Leap

Konrad Hinsen hinsen at cnrs-orleans.fr
Wed May 2 17:32:00 UTC 2007


On 02.05.2007, at 16:15, Mark Williamson wrote:

> Using the commands angle and bond I get the same number of terms as in
> the second part of the MMTK output (21 and 36), but the dihedrals  
> output
> gives a different number 52 vs 32. This is the output from rdparm:
> (justified a little for readability)

I assume that "pk" is the force constant prefactor in the dihedral  
terms. In 20 of the terms listed by RDPARAM, it is zero. These terms  
are automatically discarded by MMTK. Both AMBER and MMTK thus have 32  
non-zero dihedral energy terms, impropers included.

Note that there can be a difference between AMBER and MMTK in  
improper dihedral terms. Their number should be the same, but their  
atom orders (and thus values) can differ. The reason is that the  
AMBER force field definition for impropers depends on one non- 
physical parameter: the atom index.

For each group of four atoms that make up an improper dihedral term,  
there are three possible definitions for the angle. Instead of using  
an average of those three, which would be physically sound, AMBER  
chooses one of them based on a criterion of alphabetical order of  
atom types plus atom indices if the atom types are the same. As far  
as I know, this detail is nowhere documented; one of the AMBER  
developers explained it to me when I asked the question many years  
ago. MMTK applies the same rule as AMBER, but the atom indexing in  
MMTK and AMBER are in general not the same.

I don't see a simple fix for this, as the AMBER force field  
definition is not physically reasonable. Energies should not depend  
on the atom order in the PDB file.

Konrad.
--
---------------------------------------------------------------------
Konrad Hinsen
Centre de Biophysique Moléculaire, CNRS Orléans
Synchrotron Soleil - Division Expériences
Saint Aubin - BP 48
91192 Gif sur Yvette Cedex, France
Tel. +33-1 69 35 97 15
E-Mail: hinsen at cnrs-orleans.fr
---------------------------------------------------------------------






More information about the mmtk mailing list