Question for the experts: In a case like this with strong translational NCS, might one get value from using a least squares refinement target rather than maximum likelihood? As far as I understand it (which ain't so far) I would assume that least squares would be less influenced by the systematic variation in intensities of the reflections than maximum likelihood would. I note that phenix.refine has the option of least squares "target = *ml mlhl ml_sad ls". Can anyone comment on whether this has helped the refinement in cases with translational NCS? (I myself have been luck enough to avoid such cases thus far!) Cheers, Stephen On 4/30/08, Paul Adams wrote:

Hi,

the presence of this level of translational NCS (which you should be able to see from the arrangement of your molecules) is likely to lead to some issues with refinement. You will have sets of strong and very weak reflections. In this case you are likely to see poorer R- factors than you otherwise might expect at the same resolution (as there will be a lot of noise associated with the very weak reflections). However, you don't say what the R-factors are though.

Cheers, Paul

On Apr 29, 2008, at 2:08 PM, a a wrote:

...

Dear all: We have a set of crystal data having psuedo-translational symmetry when examined with xtriage: ---------------------------------------------------------------------- --------- Twinning and intensity statistics summary (acentric data):

Statistics independent of twin laws - /<I>^2 : 2.519 - <F>^2/ : 0.724 - <|E^2-1|> : 0.854 - <|L|>, : 0.504, 0.337 Multivariate Z score L-test: 1.854 The multivariate Z score is a quality measure of the given spread in intensities. Good to reasonable data are expected to have a Z score lower than 3.5. Large values can indicate twinning, but small values do not necessarily exclude it.

No (pseudo)merohedral twin laws were found.

Patterson analyses - Largest peak height : 67.742 (corresponding p value : 4.897e-06)

The analyses of the Patterson function reveals a significant off- origin peak that is 67.74 % of the origin peak, indicating pseudo translational symmetry. The chance of finding a peak of this or larger height by random in a structure without pseudo translational symmetry is equal to the 4.8972e-06. The detected tranlational NCS is most likely also responsible for the elevated intensity ratio. See the relevant section of the logfile for more details. The results of the L-test indicate that the intensity statistics behave as expected. No twinning is suspected.

---------------------------------------------------------------------- --------- The structure was solved by auto-sad at 2.7A. However, the r-factor can not be lowered by phenix.refine (TLS+ADP+rigid). Is there any manipuliations we can do to deal with this situation? Thanks,

Zhiyong Ren, MD Anderson Cancer Center

______________________________________________________________________ ______________ Be a better friend, newshound, and know-it-all with Yahoo! Mobile. Try it now. http:// mobile.yahoo.com/;_ylt=Ahu06i62sR8HDtDypao8Wcj9tAcJ _______________________________________________ phenixbb mailing list [email protected] http://www.phenix-online.org/mailman/listinfo/phenixbb

-- Paul Adams Deputy Division Director, Physical Biosciences Division, Lawrence Berkeley Lab Adjunct Professor, Department of Bioengineering, U.C. Berkeley Vice President for Technology, the Joint BioEnergy Institute Head, Berkeley Center for Structural Biology

Building 64, Room 248 Tel: 510-486-4225, Fax: 510-486-5909 http://cci.lbl.gov/paul

Lawrence Berkeley Laboratory 1 Cyclotron Road BLDG 64R0121 Berkeley, CA 94720, USA.

--

_______________________________________________ phenixbb mailing list [email protected] http://www.phenix-online.org/mailman/listinfo/phenixbb

-- Dr Stephen Graham Nuffield Medical Fellow Division of Structural Biology Wellcome Trust Centre for Human Genetics Roosevelt Drive Oxford OX3 7BN United Kingdom Phone: +44 1865 287 549

I hope it's OK if I chime in with this problem as well. I am working with a site-directed mutant of a commonly studied protein in our lab. It crystallize in P3121 and its unit cell is doubled along the c axis relative to a unit cell (P3221) that we have commonly seen with other mutants (the edges are roughly 55 x 55 x 88 or 55 x 55 x 176). Xtriage also picked up the pseuotranslational symmetry (0, 0, 0.5), which is not surprising given the unit cell dimensions.

If you are jusing the latest version (of of the lastes) it should actually tell you what the unit cell would be if the pseudo trans;lation would be a true crystallographic translation.

I was aware of the problem with weak and strong reflectiions, but that is not so obvious in our case. In HKLview one can see that in layers perpendicular to the c* axis, strong and weak reflections alternate out to about 8 Å, but the rest of the data (to 2.6 Å) appears - at least visibly - to be roughly consistent in intensity. Certainly when looking at the h0l or 0kl planes, there isn't any obvious alternation at higher resolution.

If the differences are subtle / relatively minor, this is what is to be expected (watch out, a lot of hand waving): at low resolution, the symmetry does not "appear to broken" resulting in strong modulations as if it were crystallographic symmetry. AT high resolution, deviations from ideality are so large that modulations are not so obvious anymore. What is the size of your off origin patterson peak as computed by xtriage?

We quickly obtained a model by molecular replacement but have stalled in refinement at an R/ Rfree of 30/38 (we have not tried TLS yet). The most obvious difficulty with the model is seen using density fit analysis in coot - one domain of one of the subunits in the asymmetric unit (related by translation NCS to the other) has weak density.

try TLS, it might help.

Maybe this is just the luck of the draw - a crystal whose contents include 50% well ordered subunits and 50% not well ordered, but I'm curious if there is something obvious we should be trying.

Not sure. 50% ordered and 50% not ordered sounds a lot like order disorder twins (OD). I am not sure what canb by done with that type of data in terms of refinement.

By the way, xtriage found no evidence of twinning, and the data also passed muster with dataman (Padilla-Yeates) and Todd Yeates' server.

Xtriage does use the Padilla Yeates test as the primairy vehicle for twin detemination, combined with the presence of twin laws of course. Beware that the presence of a full complement of nightmares such as OD twinning, merohedral twinning, pseudo symmetry etc etc can make intensity statistics look like all is fine. P

Thanks in advance,

Arthur Glasfeld Department of Chemistry Reed College

P.S. If anyone is interested I can send a figure with the hklview layers perpendicular to c* (hk0, hk1, hk2, hk3, hk4, hk5)

_______________________________________________ phenixbb mailing list [email protected] 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 -----------------------------------------------------------------