Scalepack

[Zbyszek Otwinowski]

Serena Cooper wrote:
> 
> I'm having some problems using scalepack. I have 2 separate data sets
> (synchrotron) from two different  crystals, both tetragonal one to 2.5A another
> 2.8A. Indexing with Denzo goes OK. When I postrefine the 2.8 data set the cell
> dimensions are all over the place often decreasing by up to -7A and similarly
> with the mosaicity, which gives nonsense values. I then get a list of
> parameters that it cannot refine. If I split the postrefinement up into blocks
> of 3-10 images then everything settles down - could this be due to slippage??
> The 2.5 data set behaves slightly differently - if I do not postrefine I get
> resonalble chi2 and R-merges. Post refinement again gives fluctuating cell
> parameters (+-4A) and the mosaicity jumps from ~0.6 to >1.5.
> Is this just bad data , or ...
> 

The postrefine option in Scalepack is very flexible.
However, this flexibility may be misused.
1.)
If you have 2 crystals, you have to decide if you assume that
they have the same unit cell in postrefinement, or that
you refine the unic cell separately.
2.)
Under most circumstances it is recomended that crystal
orientation is refined separately for every batch (image) of data.
You should refine

fit batch rotx 5 to 152
fit batch roty 5 to 152

(if your range of batch numbers is 5 to 152)

You must not

fit batch rotz

as this may lead to unstable refinement.
rotz is rotation of crystal around beam axis.
This rotation does not affect noticably prediction
of reflections partiality. Postrefinement is bases
on comparing predicted and observed partiality.
If prediction is not dependent on particular
parameter, refinement is unstable.

3.)
In P4 (and P41, P42, P43), two crystals may have
inconsistent indexing.

> When I postrefine the 2.8 data set the cell
> dimensions are all over the place

Postrefinement of the unit cell separately on each
batch is not generally recomended. There is simply
not enough information there. So observed behaviour
is as expected in postrefinement procedure.

> If I split the postrefinement up into blocks
> of 3-10 images then everything settles down - could this be due to slippage

Blocks of 3-10 images refine more stably than single images.
Even better, in most cases, the same value of unit cell may assumed
for all the batches from one crystal.
Refinement of mosaicity in groups of 3-10 batches is sensible
if mosaicity increases due to radiation damage or if mosaicity
is very anisotropic.

> Post refinement again gives fluctuating cell
> parameters (+-4A) and the mosaicity jumps from ~0.6 to >1.5.

This is probably due to bad setup of postrefinement.
-- 

General comment:

Usually is best to start postrefinement by coping an appropriate
example (scenario) from scalepack manual.

--
  Zbyszek Otwinowski                     |        zbyszek at chop.swmed.edu
  University of Texas                    |        tel : (214)-648-5098 
  Southwestern Medical Center            |        fax : (214)-648-5095 
  5323 Harry Hines Boulevard  Dallas, Texas  75235-9038
[EOB]