[gmx-users] OPLS united atom

[Erik Lindahl]

Krzysztof Murzyn wrote:
> Hi,
> 
> I'd like to transfer OPLS united atom forcefield as implemented in Amber
> package (files: opls_parm.dat and aa residues definitions) into the
> Gromacs format. The first problem I faced concerns dihedrals.
> 
> The Amber definitions give 1/2 of the rotation barrier magnitude (in units
> kcal/mol), periodicity and the phase (0 deg if an energy MAXIMUM is at 0
> deg or 180 deg if an energy MINIMUM is at 0 deg; 0 deg corresponds to
> the cis conformation). This form matches (?) Gromacs definition with the
> function form eq 1. Amber definitions of some dihedrals consist of a few
> terms resulting from deconvolution of the full dihedral potential into a
> Fourier series. Each term is given in a form described above and it
> happend that terms for a single dihedral have a different phase.
> 
> My question is how I could properly transfer these parameters in the
> Gromacs world. An obvious solution would be to list all terms for each
> torsion with func eq 1 but this implies that one Gromacs dihedral
> definition would consist of several entries/lines. Would it be OK?
> 

No, rewrite it as a Ryckaert-Bellemans dihedrals instead; this is 
described in the manual, and I've already done it for OPLS-AA.

The reason why we don't use several dihedrals for one bond is that you 
would end up calculating cos(phi) two or three times, and this is a 
somewhat costly operation. With a series we can calculate it once and
then just multiply it to get higher powers.

Cheers,

Erik