SHELX does not need a parameter to control the relative weights of
restraints and X-ray data, it is performed automatically. However you
can change the restraint esds, either individually or globally using
DEFS, though the default values are generally OK. There is bound to be
some difference in R-values between the two programs, e.g. because of
differences in the bulk solvent model. In general there will also be a
tendancy for least-squares refinement programs like SHELXL to give a
bigger difference between R and Rfree than maximum likelihood programs
like REFMAC, especially at modest or low resolution (which is the main
reason why we also use REFMAC when the resolution is worse than about 1.8A).
For very large structures it is necessary to split SIMU and DELU
instructions into several instructions, which requires understanding how
SHELXL references atoms (this is very flexible but still intuitive, but
I'm afraid that there is no alternative to reading the MANUAL!). You
will not be needing DELU at 2A anyway, it is only required for
anisotropic refinement.
For refinement of the disordered ligand, if there are two conformations
you can use a single 'free variable' so that one conformation has the
occupancy p and the other 1-p. Alternatively, if there is more than one
conformation or the occupancies do not add up to one, just use one free
variable per ligand to describe the occupancies. The flexible treatment
of disorder is the main reason (together with twinning) why we prefer to
refine such structures wiht SHELX despite the advantages of maximum
likelihood. On the other hand TLS refinement in REFMAC is sometimes very
effective.
As well as the manual, I should mention that there will be a SHELX
WORKSHOP on the first day of this year's ACA Meeting at Salt Lake City.
And Martyn Winn has just finished patching mtz2various so that you can
at last move the reflection data from the mtz file to the SHELX hkl file
without the excruiating contortions that were previously required.
However you will need to download and recompile the patched version
yourself, and it is not yet in CCP4i.
George Sheldrick
Artur Matysik wrote:
> Dear CCP4BB users;
>
> Due to mixed state of a ligand conformation I wanted to use
> shelx, to refine it correctly at the end point of refinement.
>
> My problem is, that I do not know how to control geometry in Shelx.
> After Shelx refinement, my bonds are very tight (rms 0.008), angles
> loose (rms ~1.9), R is lower (21.2) and Rfree (28.6) is much higher,
> comparing to Refmac results (R/Rf 22.9/27.6 rms Bond/Angle 0.011/1.1561)
> (statistics calculated doing 0 Refmac cycles on Shelx refined model,
> res 2A, refinement without water molecules).
>
> Is there an equivalent of Refmac's 'weighting' in Shelx?
> How to control X-ray vs geometry ratio?
> I have tried with different WGHT parameter, but in the best case
> there is no change in rms bonds and little in angles,
> Rf is getting better (to 27.9), while R drops a bit (20.2) -
> overfitting? .
> Different FVAR also make no big difference.
> How to deal with that?
>
> and my second question:
> Shelx complains that that I have too many atoms on a single SIMU/DELU
> instructions.
> Following FAQ, I should 'split the offending DELU (and possibly SIMU)
> instructions'
>
> f.ex.
> DELU N_1001> OT2_9999
> DELU O_1 > O_999
>
> but I am still not sure, what to feed Shelx instead of default:
>
> DELU $C_* $N_* $O_* $S_*
>
> Thanks for any help : )
> Art.
|