[phenixbb] xtriage to double-check symmetry

Peter Zwart phzwart at gmail.com
Thu Sep 17 20:40:05 PDT 2009

Hi Frank,

There are two types of space group problems:

1. Too much assumed symmetry
2. Not enough assumed symmetry

Category 1 is nasty: the structure doesn't refine very well or doesn't
solve at all. This is due to the fact that the 'extra' symmetry you
have introduced act as constraints on the solution. These symmetry
constraints are in fact incorrect, and unless you remove them, things
will be not be "im gruenen bereich".

Category 2 is a better place to be in. In that case, you can do the
RvsR test and inspect the table. If you have an R value for calculated
data that is equal to (or very close to) 0, that almost always means
that you have missed rotation/screw symmetry somewhere. If on the
other hand Robs is near zero, but Rcalc isn't, your data is perfectly
twinned. In fact, any significant reduction in Robs with respect to
Rcalc, is typically due to twinning.

When you miss symmetry, you  typically do not notice, unless you look
for it. This is due to the fact that the constraints of space group
symmetry are turned into something like restraints imposed by inherent
symmetry in the data. Turning a constraint in a restraint is typically
not very bad, but making a restraint in a constraint, typically makes
life difficult.

So in short, the only way to find out what's happening, is to remove
symmetry operators from your high symmetry solution (i.e. MR in a
lower space group), solve and analyze the results.

Xtriage can only make specific suggestions which symmetry to add to
your model. I do not see how I can make a suggestion which elements to
remove without doing the extra refinement steps in lower symmetry
space groups.

Does this help?


2009/9/17 Frank von Delft <frank.vondelft at sgc.ox.ac.uk>:
> Then I'm afraid I've totally misunderstood what this does, and I don't
> see it in the docs either.
> It *sounds* like it would take your data and model, and check whether
> going to higher symmetry would give as good an agreement.
> That's obviously not what it does; could you clarify?
> Thanks!
> phx
> Peter Zwart wrote:
>> Hi Frank,
>> You can use P65 as well. IN p6522 there are no extra operators
>> (typically),  so not much can be learned (the r value for calculated
>> data will be zero). You do need to have lower symmetry.
>> HTH
>> Peter
>> 2009/9/17 Frank von Delft <frank.vondelft at sgc.ox.ac.uk>:
>>> Hi
>>> We were trying to decide whether a structure is P65 or P6522:  data
>>> merges fine in latter, and NCS is perfect, but R/Rfree are significantly
>>> worse (32/37-ish vs 23/28 for P65).
>>> So I thought I'd listen to xtriage:  seemed easy since I have both data
>>> and model in P6522, and I ran:
>>>     phenix.xtriage final.mtz reference.structure.file=final.pdb
>>> I'm not sure it's done anything, though - at least, I can't find the
>>> table shown on the 4th-last slide of this presentation:
>>>    http://aca.hwi.buffalo.edu/HotNews/twinning/zwart.pdf
>>> Something I said / did not say?  Do I *have* to run it in P1?
>>> Thanks!
>>> phx
>>> _______________________________________________
>>> phenixbb mailing list
>>> phenixbb at phenix-online.org
>>> http://www.phenix-online.org/mailman/listinfo/phenixbb
> _______________________________________________
> phenixbb mailing list
> phenixbb at phenix-online.org
> http://www.phenix-online.org/mailman/listinfo/phenixbb

P.H. Zwart
Beamline Scientist
Berkeley Center for Structural Biology
Lawrence Berkeley National Laboratories
1 Cyclotron Road, Berkeley, CA-94703, USA
Cell: 510 289 9246
BCSB:     http://bcsb.als.lbl.gov
PHENIX: http://www.phenix-online.org
CCTBX:  http://cctbx.sf.net

More information about the phenixbb mailing list