Re: [ccp4bb]: REFMAC dictionaries etc

[Ethan Merritt <merritt@u.washington.edu>]

***  For details on how to be removed from this list visit the  ***
***          CCP4 home page http://www.ccp4.ac.uk         ***

On Tuesday 16 October 2001 03:39, you wrote:
> We are refining a structure at 1.54A resolution.
> We know we have NADH bound and the original Fo-Fc density shows the ring is
> distorted into a boat.
> Refining B's isotropically ends up with the ring in a twist boat
> conformation (even although it is planar restrained). Obviously this is
> well and good. So I reconstructed the dictionary by removing planar
> restraint and adjusting some bond lengths and angles. Dihedral angles I
> widened the restraints for. The result is the NADH now becomes a chair
> (bad).
>
> Also anisotropic refinement causes the ring to flatten out with my new
> slack restraints.

That actually sounds reasonable to me.  Out-of-plane density around the
ring is now being accounted for by the ellipsoidal atom descriptions, rather
than by displacing the atomic centers.  That doesn't mean this model is
necessarily correct,  but the behaviour of the refinement program is
plausible.

> Any thoughts welcomed!

It sounds like you took the entire model anisotropic all at once, including
the NADH.  Is that correct?  I wouldn't do that.  I'd take the protein aniso
and refine it to get better difference density maps (mFo-Fc) for the region 
of the NADH. 

If the conformation of the NADH remains ambiguous, I would then try to
formulate a specific statistical test to descriminate between the possibilitie.
My first thought is to make 3 parallel runs, one with the NADH geometry
tightly restrained to boat, one with twist boat, one with chair.  Your
choice of iso or aniso treatment of the NADH atoms, or try both (6 runs).
A ligand is often less well ordered than the protein itself, and may well
be handled best by keeping it's model isotropic even after taking the
protein aniso.

You may or may not see a difference in Rfree, but really I would be 
looking for the plausibility of local measures like residual density
(quantified by DDQ perhaps), rmsd of heighboring B factors (isotropic
case), rmsd violation of DELU restraint (anisotropic case), plausibilty
of ortep/rastep plot showing implied vibrational modes. 

> The distortion in the ring
> will be small, could it be soaked up by error in the ellipsoids?.

Not sure what you mean by "error in the ellipsoids".  The ellipsoids
constitute a model for the local density, and if there is error (noise)
in the local density then the ellipsoids will try to descrube it.  Is that
what you mean?

-- 
Ethan A Merritt       merritt@u.washington.edu
Biomolecular Structure Center Box 357742
University of Washington, Seattle, WA 98195
phone: (206)543-1421
FAX:   (206)685-7002