 |
 |
Hi Sudipta,
The combination of crystallographic symmetry, NCS and tNCS is extremely tricky and several things may explain the somewhat high Rfactors:
1)
One of the 2-folds you thought is crystallographic is in fact non-crystallographic, e.g. a few Å off from its crystallographic position.
You could try to expand your data to P1 and run a rigid body refinement of your structure at low resolution (4-4.5 Å) to see what happens. Alternatively, you could run MR in P1 using your refined model.
2)
In case of tNCS, you have two layers of molecules, say A and B, which are almost, but not 100% identical (otherwise the translation would
be crystallographic). During refinement, you might not have been able to break the symmetry between the A and B layers. You could try to run a number of refinement cycles without imposing NCS and look in the maps to see if differences show up.
3)
Instead of neatly packing ABABABAB you might have something like ABABAABABABBABAB
and this disorder might be behind your high Rfactors. This might show up as lattice translocation disorder, which may lead to some of your diffraction spots being fuzzy, while other are sharp. In your case, I would check the diffraction images for the presence
of fuzzy spots.
4)
It might be an artifact of half your diffraction spots being (very) weak. In Sftools, it should be possible to select only the strong
reflections and calculate Rfactors for them. If those Rfactors are ok, you could decide that your refinement is finished and the structure is ok. If your weak reflections are very weak, and upon superposition of the A and B layers, there are no significant
differences, you could also decide to ignore the tNCS and process and refine the structure using half the unit cell axis. The resulting electron density will be the sum of the A and B layers.
Again, your crystal form is extremely tricky and you may have to live with the high Rfactors when it is not possible to adequately model the (unknown)
disorder.
Best,
Herman
Von: CCP4 bulletin board [mailto:[log in to unmask]]
Im Auftrag von Sudipta Bhattacharyya
Gesendet: Samstag, 11. Juli 2015 18:14
An: [log in to unmask]
Betreff: Re: [ccp4bb] refmac5/phenix refinement strategies with 3.2A data with pseudotranslation.
Hi Tony,
Thanks for your input. However, I am almost towards the end of model building...need to join some loops and so..and that was the R/Rfree...The Scala/Xtriage/Ctruncate output did not show any signs of twining...although in presence of tNCS these tests are all
blurred....but even in phaser MR which detected the tNCS vector and corrected it eventually, did not show the presence of twining...I think even without the presence of twining, twin refinement is not a good option to turn on. Please let me know what you think.
Best,
Sudipta.
On Sat, Jul 11, 2015 at 5:39 AM, Antonio Ariza <[log in to unmask]> wrote:
Hi Sudipta,
Your R factors don't look horribly high for an initial build and refinement at 3.2A, assuming you still need to substantially improve your structure ...
if your structure is basically complete that's obviously a different matter. First of all I would look at your scaling output to check for twinning (aimless and pointless give very good indications) ... or simply re-run your last refmac job with twin refinement
"on" and check the log file to see if the programme finds any twinning.
Regards,
Tony
------------------------------------------------------
Dr. Antonio Ariza
University of Oxford
Sir William Dunn School of Pathology
South Parks Road
Oxford
OX1 3RE
From: CCP4 bulletin board [[log in to unmask]]
on behalf of Sudipta Bhattacharyya [[log in to unmask]]
Sent: 10 July 2015 19:59
To: [log in to unmask]
Subject: [ccp4bb] refmac5/phenix refinement strategies with 3.2A data with pseudotranslation.
Dear All,
I am struggling with a 3.2A structure refinement which also suffers from translational NCS issue. Although the maps look very good but R/Rfree is not going down as expected. Could someone
please suggest me tricks/tips with Refmac5/Phenix refinement optimization in this regard. The model has 4 molecules/ASU, space group initially appeared P212121 but later was found to be P21212; current R/Rfree is 33/36. Please let me know if you need any further
details. Any suggestion/help will be really appreciated.
My best regards,
Sudipta.