Bernhard,
I think for the loop that is about to close taking the average from
flanking residues is good enough. Not sure what you mean by the second
example ("last residue has a low B").
I think what causes problems in refmac is a sharp discontinuity, as it
contributes greatly to the B-factor restraints. Refinement will adjust
B-factors properly if they are higher/lower, but not if they jump from
90 to 20 from residue to residue.
The way I adjust B manually is by looking at the B-factor of the next
residue, which is not too sophisticated approach and should be easy to
implement. Well, in the end it's a minor problem, and easy to fix
manually.
Cheers,
Ed.
On Thu, 2010-11-04 at 16:11 +0100, Bernhard Lohkamp wrote:
> Dear Ed,
>
> I think I get your idea, however this is not a straight forward issue I
> believe. What do you imagine happens for a loop then? You may end up
> connecting a high B residue to a low one when closing it? Does that make
> refmac happy? Or your last residue (since it is still well ordered) has
> a low B, then you dont gain anything? More appropriate may be a
> weighting and adjustment of B-factors based on e.g. real-space CC. But I
> guess until we have tools in Coot to refine (or in the first instance
> guess) B-factors you better adjust the Bs yourself. Your guess (knowing
> the structure and seeing the map) is better than any approximation we
> can easily come up with (at least for now).
>
> My 50 Oere (just fazed out BTW),
>
> Bernhard
>
> > When adding residues to the N- or C-terminus, it may be desirable that
> > the B-factors of the new residue match that of the previous terminal
> > residue. Right now it uses defaults (20 for main chain, 30 for side
> > chain) which are sensible fir a well-ordered model, but this causes
> > problems with the subsequent refmac run for somewhat disordered regions
> > where<B> may be significantly higher.
> >
> > The current workaround is to modify residue properties manually (or keep
> > resetting the defaults). Naturally, the wisdom of adding extra residues
> > with<B>~100 is questionable :)
> >
> > Cheers,
> >
> > Ed.
> >
>
>
--
Coot verendus est
