[gmx-users] amber96 RB dihedrals in GMX 4.0 and 4.5

[Mark Abraham]

On 25/01/2011 4:53 AM, Badi' Abdul-Wahid wrote:
> Hello
>
>
> Context:
> I have prepared the same system (WW domain) with GMX 4.0.7 and
> 4.5.3 using amber96. Doing a gmxdump on the tpr files seems to
> show that the RB dihedrals are absent from the 4.5-prepared
> system, but present in the 4.0-prepared one. I used the same mdp
> files for both GMX versions.
>
> > gmxdump -s system.tpr | grep RBDI
>
> Looking at the [dihedrals] directive in the topology file: 4.0
> has funct 1 (proper dihedrals) and funct 3 (RB dihedrals) while
> 4.5 has funct 4 (improper) and funct 9 (proper) (pages 108 and
> 124 for the 4.0 and 4.5 respectively).

Building AMBER topologies requires that the residue database have the 
ability to add up multiple instances of type 1 dihedrals, which is an 
ability that was lacking in 4.0. Type 9 was introduced in 4.5 to deal 
with this and similar situations. Under some circumstances, multiple 
type 1 dihedrals can be combined into an R-B dihedral, and this was the 
work-around in 4.0. Function type 9 avoids this need. You should find 
that each produces the same energy in a zero-step MD. I would have 
expected the improper dihedral treatment in each GROMACS version to be 
the same.

>
> The 4.5 manual (page 129 under the [dihedrals] directive
> description) seems to imply that the RB dihedrals need to be
> added by hand and the 1-4 interaction in [pairs] should be
> removed for the atoms involved in the RB dihedrals.
> Is this correct?

I think not. This is force-field- and system-specific. See 4.2.12.

>
>
> Question:
> What is the best way to prepare a system with Ryckaert-Bellemans 
> dihedrals in GMX 4.5?

I think you should have no need to use them with amber96.

Mark