I'm not a big fan of restraining chiral volumes, except for models based
on low resolution data (maybe up to 2.7A). In my experience, if a refinement
results in the flipping of a chiral center the true solution is almost never
to flip it back. If a CA flips then probably a neighboring peptide bond
needs to be flipped. If a threonine CB flips then the OG1 and CG2 labels
need to be swapped.
I would prefer a visual indicator on the screen when a chiral center is
bad in addition to your tool for jumping directly to bad centers. That
way, when I'm passing down the chain I would see, say, an icon with a
picture of a hand with a red circle and diagonal slash and know immediately
that there is a problem. Of course I would also like to see visually the
bad bond lengths and angles, but marking the bad chiral centers would not
clutter the screen and be a good start.
Dale Tronrud
Paul Emsley wrote:
> On Tue, 2007-03-20 at 08:57 +0100, Laurent Maveyraud wrote:
>> Dear list,
>>
>> another question related to stereochemistry, but concerning coot.
>> When using the RSR zone button, which should "improve geometry and fit to
>> map", it seems that the geometry is not always improved : some chiral
>> centers can be inverted.
>
> That is not always a bad thing.
>
>> The surprising thing is that using the "regularize
>> zone" afterwards does not correct this, even if the chirality of the center
>> is not set to both, but to positive in the cif dictionary (ie should define
>> only one possible chirality).
>> Has anybody else notices this behavior ? Is there a way to adjust the weight
>> between geometry and map terms in the RSR option ?
>
> Chiral volumes are not part of the geometric target function in Coot[1],
> so changing the weight will have no effect.
>
> Currently what I do for chiral volumes (as a work-around) is use
> Validate -> Incorrect Chiral Volumes to find the building errors. They
> are trivial to fix when you know where they are. Often I use single
> atom drag to move the chiral centre to the other side of the chiral
> plane (made by the chiral neighbours).
>
> Paul.
>
> [1] They should be, I tried it but it made the minimization slow and
> unstable. I should try it again.
|