[gmx-users] How to implement dihedral restraints

[David Mobley]

Hi,

> The test I ran was just rotating a side chain of an amino acid residue
> over one of its dihedral angle 360 degrees with dihre_fc = 8000.

OK. I was doing a val sidechain.

> I originally thought fc = 8000kJ/mol unit^2 would give fluctuation of
> about the half the unit size since I use fc = 8000kJ/mol nm^2 for 0.5A spacing.
> But, after I thought twice about it, it would be fc = 80kJ/moml nm^2, for 0.5nm spacing...
> So, after I realized that, kJ/mol rad^2 made sense to me and my test run's fluctuation
> (about several degress = 1/20 rad = about 3 degrees)

OK. I was doing my valine sidechain with about 26 umbrellas using
force constants ranging from around 100 kJ/mol*rad^2 to 500
kJ/mol*rad^2 (bigger force constants when the PMF is steep). I wish I
had an adaptive scheme for picking force constants, but I just did
this by trial and error to get good overlap with as small a number of
umbrellas as possible.

I guess the other way is what it sounds like you're talking about
doing -- just put umbrellas with tight springs every couple degrees.

It is probably important to make a plot of your binned dihedral angles
at the end to make sure you have sufficient overlap.

David