[gmx-users] CHARMM force field implementation in Gromacs :

[Mark Abraham]

Nicolas SAPAY wrote:
> Thanks for your answers (I had forgotten this comment in the script).
> 
> The problem is that most of dihedral with multiplicity n >= 6 don't come
> alone. For exemple in Arg :
> 
>         HD1 HE
>     |    |   |    //
>   --CG--CD--NE--CZ
>     |    |         \
>         HD2
> 
> is defined by 6 dihedral with n>=6 (CG-CD-NE-HE, ..., HD2-CD-NE-CZ)> They
> are all of the same type (X-CT2-CT2-X). So, if I understand well, the
> result should be OK since, for example, CG-CD-NE-HE and CG-CD-NE-CZ are
> not a combination of different type of dihedrals with n >= 6.

Correct. A dihedral is defined by the four atoms, the functional form 
and the parameters. The energy profile of rotation about the CD-NE bond 
is a linear combination of however many dihedrals it is, which is a 
quite different thing.

> The problems happen when a combination of different types of dihedral are
> used (for example if CG-CD-NE-HE is of type A and CG-CD-NE-CZ is of type
> B). In this case, one possibility is to hack the Gromacs code by allowing
> a 6th Ryckaert-Bellemans parameter (?)

This is not a problem. None of these dihedral functions occupy the same 
"space", in that (by the above definition) they have a different set of 
four atoms despite having the same functional form and central atoms.

The potential problem comes because the [ dihedraltypes ] section of 
either the .top file (or the *bon.itp file that is #included) does not 
allow multiple dihedral functions of the same functional form to apply 
to the same set of four atom types - grompp just uses the last one 
defined. As David correctly points out (and I learned something here!), 
you can define multiple dihedrals of the same form for the same set of 
four *atoms* (not atom types) in the [ dihedrals ] section of the .top 
file. You have to read chapter 5 closely to pick that up though (5.6.1 
and 5.7.2). So yet another solution to the general problem would be for 
a script to post-process the CHARMM .top file to "correct" the dihedrals 
where multiple periodic functions apply to given set of four atom types.

Mark