Hi -
What Ron says, is of course correct, you should refine your structure
in the highest symmetry space group consistent with the experimental
data.
The question though sometimes is exactly what is the highest symmetry
space group consistent with the experimental data.
Although I am far from an expert in this, I faced the problem enough
times lately to allow myself to suggest a strategy, or if you prefer
summarize what I do with our new crystals ...
For a new crystal that has not be characterized, always first process
data in P1. Programs like XDS always produce the P1 data while in
moslflm its trivial to process in P1.
Programs like phenix.xrefine from phenix/cctbx and Pointless from
CCP4, provide excellent tools to help you decide what the 'correct'
symmetry is.
They will apply symmetry elements that are consistent with the metrics
of the P1 lattice (eg if all three angles are close to 90,
they will try all three monoclinic and the orthorhombic choice). These
can help you decide which symmetry elements are 'true' and which are
not;
both programs make suggestions for the 'most likely' space group.
Once the space group is known, scale and merge your data in that space
group, my preferred option is Scala here. Then go back
to .xtriage to look for twining and peaks in the native Patterson, to
exclude twining and make sure there are no non-crystallographic
translation axes.
If you ended up with a high symmetry space group, but Rfree strays
high in refinement, there is still a chance that you might have done
something wrong in the steps before (its more likely that you did not
do the steps before, but never mind). The two things I like to check
at this stage, is 1. add the TWIN keyword in refmac and let refmac
find if there is some twining possible and see if the twin fraction
refines to some reasonable value (or use .xtriage and phenix.refine to
do the same) 2. USe the BALBES server in Garib's home page. This will
'expand' your reflections to P1 and all space groups between P1 and
your assigned symmetry. Then it will try and refine the structure in
all these space groups. If the Rfree is significantly lower in one of
the lower symmetry space groups it will report it and you should
consider using that symmetry.
These still leave the choice of lower symmetry space group *and*
twining, which BALBES does not check yet (but Garib and team are busy
with this). For that I guess you still need to check manually if
refinement is stack in high Rfree while the maps looks good and the
molecule is reasonably complete.
Hope this helps.
Oops - what was my new year resolution? No lengthy emails to ccp4bb ?
Damn.
Happy new year all -
A.
On 5 Jan 2009, at 7:57, Ronald E Stenkamp wrote:
> Hi.
>
> If the most precise and accurate description of your crystal
> structure is the orthorhombic one, you should average the replicated
> reflections and refine the structure in C2221. If the crystals are
> orthorhombic and you refine and report it in C2, you're producing a
> model that is larger and more complex than you need to explain the
> experimental data. You're also reporting to the world that the two
> molecules in the C2 asymmetric unit are different when there's no
> experimental evidence supporting that.
>
> To extract the most information from your diffraction data, you
> should refine your structure in the highest symmetry space group
> consistent with the experimental data.
>
> Ron
>
>
> On Sat, 3 Jan 2009, Yu Jiang wrote:
>
>> Dear all,
>>
>> I am now refining one structure in space group C2 with 2 molecules,
>> however
>> we find the data can be processed in C2221 with 1 molecule. The
>> questions
>> are:
>>
>> 1. Can I still use space group C2 instead of C2221, even if the two
>> molecules almost have crystallographic symmetry? Should I provide any
>> excuses?
>>
>> 2. Is there any indication or necessity to refine structures in
>> lower space
>> group instead of higher?
>>
>> Many thanks!
>>
>>
>>
>>
|