[gmx-users] Dihedral angle restraints in NMR refinement...

[Erik Lindahl]

Chris Spronk wrote:

> Hi there,
>
> I have been playing a little bit with the NMR derived dihedral angle
> restraints as defined as impropers. I ran into the following problem: 
> Dihedral angle restraints close to 180 degrees give rise to local
> heating and a lot of LINCS warnings (and exploding system).
>
> On page 59 of the gromacs manual it is stated that impropers should be
> defined as far away from 180 degrees as possible. The reason (dank je
> Bert!) is that the periodicity is absent and small changes in the
> dihedral angle leading to flips from ~-180 to ~+180 lead to huge
> differences in improper dihedral energy.
> So when I converted a couple of the dihedrals that were close to 180 ( N
> CA CB CG ) to dihedrals around ~60 degrees ( C CA CB CG ) these
> restraints did not give rise to any LINCS warnings or heating anymore
> (but a little later some others around 170 degrees did).
> Although this is a possible work-around it still makes the use of
> improper dihedral angles for dihedral angle restraints rather 
> impractical. In addition, I think it limits the sampling of 
> conformations.
>
> So I was hoping someone of the developers-team could help us (the NMR
> users) with this. In addition it would still be nice to have a restraint
> type similar to the distance restraints, with zero force between an
> upper and lower bound for the angle to account for the experimental 
> uncertainty, and a harmonic potential outside
> these bounds. I hope this is not too big of a problem, but otherwise NMR
> people will probably not be prompted to use GROMACS and I personally
> think that would be a great pity....
>
> thanks in advance,
>
> Chris

Hi Chris,

I'm sure we can fix this; adding bonded interactions like constraints is 
trivial, and if we only determine some reasonable form for the
interaction I promise to implement it. Do you know how it's done in 
other programs?

Cheers,

Erik