Re: [ccp4bb] R factor R free struck
As always, look at the crystal packing of the MR solutions. That is a quick way to reject incorrect space group solutions. I never realized Phaser had the issue you described with the SGALTERNATIVE setting, but to be honest, I've always run it in "manual" mode, i.e., I set up separate jobs in all the alternative space groups and check Z-scores and packing. It's a bear in P222 or C222 with all the possible re-indexed alternatives, but it works. It's only CPU time. Cheers. Steven Sheriff wrote: How certain are you of the space group? I ask because in all versions of PHASER that I have used to date, it chooses the space group (when SGALTERNATIVE is set) after placing the first molecule and with additional molecules it then assumes the space group result from the first molecule is the correct space group (the PHASER team is supposed to be rectifying this). While this is often the correct space group, I have seen several cases with both PHASER and other molecular replacement programs (e.g. AMoRe) where the correct space group only becomes evident after the placement of the second molecule. In particular, I have seen a case where after the first molecule was placed, one would have chosen P 21 21 2 on the basis of statistics provided, but after the second molecule was placed, it was clearly P 21 21 21. Steven -- Roger S. Rowlett Professor Department of Chemistry Colgate University 13 Oak Drive Hamilton, NY 13346 tel: (315)-228-7245 ofc: (315)-228-7395 fax: (315)-228-7935 email: rrowl...@mail.colgate.edu
Re: [ccp4bb] R factor R free struck
On Sun, Oct 3, 2010 at 6:34 PM, J. Preben Morth pr...@bioxray.au.dk wrote: hi remember to reindex your data to P21212 in case you used Phaser to search all alternative orthorhombic SG's and it found P22121 Preben Hi, re-indexing the data without also remembering to transform the co-ordinates (using the matrix transpose of the inverse of the re-indexing operator) would almost certainly cause the R factors to increase! But even assuming you did that, why should re-indexing/transformation have any effect at all on the R factors? Cheers -- Ian
Re: [ccp4bb] R factor R free struck
Is that necessary? Eleanor J. Preben Morth wrote: hi remember to reindex your data to P21212 in case you used Phaser to search all alternative orthorhombic SG's and it found P22121 Preben On 03/10/2010, at 04.56, Jack Russel wrote: Hi all, I have collected a data at 2.9 Å and the solved the structure using phaser . the space group comes to be P2 21 21. There are 4 molecules in Assymetric unit and an octamer is generated according to the symmetry. But after repeated rounds of rigid body refinement with REFMAC5 and model building with coot the R factor had been struck at 40% and R free at 50%. So my first question is whether my solution after phaser is correct. And if it is how can i lower the R factor and Rfree. The second question is it possible to have such a large difference between R factor and R free. Thanks in advance
Re: [ccp4bb] R factor R free struck
the reindexing have no effect on the R-factor, I did indeed mean to remember to transform the co-ordiantes according to the indexing :-) On 04/10/2010, at 11.27, Ian Tickle wrote: On Sun, Oct 3, 2010 at 6:34 PM, J. Preben Morth pr...@bioxray.au.dk wrote: hi remember to reindex your data to P21212 in case you used Phaser to search all alternative orthorhombic SG's and it found P22121 Preben Hi, re-indexing the data without also remembering to transform the co-ordinates (using the matrix transpose of the inverse of the re-indexing operator) would almost certainly cause the R factors to increase! But even assuming you did that, why should re-indexing/transformation have any effect at all on the R factors? Cheers -- Ian Jens Preben Morth, Ph.D Aarhus University Department of Molecular Biology Gustav Wieds Vej 10 C DK - 8000 Aarhus C Tel. +45 8942 5257, Fax. +45 8612 3178 j...@mb.au.dk website: http://person.au.dk/en/j...@mb.au.dk
Re: [ccp4bb] R factor R free struck
Try a lower symmetry, e.g. P21 or P1 with one or two octamers in the asymmetric unit (well, unit cell for P1), but you already know your packing so no problem solving it. It can sometimes be very subtle and your model refinement will be the most sensitive test for the correct space group - so don't be misled by seeming perfect scaling in orthorhombic. Poul On 03/10/2010, at 04.56, Jack Russel wrote: Hi all, I have collected a data at 2.9 Å and the solved the structure using phaser . the space group comes to be P2 21 21. There are 4 molecules in Assymetric unit and an octamer is generated according to the symmetry. But after repeated rounds of rigid body refinement with REFMAC5 and model building with coot the R factor had been struck at 40% and R free at 50%. So my first question is whether my solution after phaser is correct. And if it is how can i lower the R factor and Rfree. The second question is it possible to have such a large difference between R factor and R free. Thanks in advance
Re: [ccp4bb] R factor R free struck
A handy way to do what Paul suggests is the Zanuda server at York which will try refinement in all lower symmetries ... From the first email though two things are unclear: 1. How good was the Phaser solution (eg z-score?) ? 2. Did you only try rigid body fitting as you wrote? (I find regid body after Phaser frankly useless) Anyway, what I would do would be to: a. get the latest version of refmac before doing anything else, and do full XYZB refinement, not rigid body, with tight NCS restraints b. if that does not drop Rfree to say below 45%, submit to Zanuda to check symmetry as Poul suggests - do that anyway in fact! c. if zanuda gives lower symmetry, reprocess your data in lower symmetry, run Phaser again, run refmac again, check for twining Maybe I will go now on preaching the wrong gospel for this BB, but Buster and/or Phenix are extremely useful at that resolution range. I must say that Buster sometimes has been really impressive in our hands. A. On 3 Oct 2010, at 9:09, Poul Nissen wrote: Try a lower symmetry, e.g. P21 or P1 with one or two octamers in the asymmetric unit (well, unit cell for P1), but you already know your packing so no problem solving it. It can sometimes be very subtle and your model refinement will be the most sensitive test for the correct space group - so don't be misled by seeming perfect scaling in orthorhombic. Poul On 03/10/2010, at 04.56, Jack Russel wrote: Hi all, I have collected a data at 2.9 Å and the solved the structure using phaser . the space group comes to be P2 21 21. There are 4 molecules in Assymetric unit and an octamer is generated according to the symmetry. But after repeated rounds of rigid body refinement with REFMAC5 and model building with coot the R factor had been struck at 40% and R free at 50%. So my first question is whether my solution after phaser is correct. And if it is how can i lower the R factor and Rfree. The second question is it possible to have such a large difference between R factor and R free. Thanks in advance
Re: [ccp4bb] R factor R free struck
1) Rigid body refinement wont reduce R factors much more than this - start restrained refinement with NC restraints.. 2) And yes - at low resolution you could expect a large difference between R and rfree 3) With such high NC symmetry is there any possibility of another SG? Eleanor Jack Russel wrote: Hi all, I have collected a data at 2.9 Å and the solved the structure using phaser . the space group comes to be P2 21 21. There are 4 molecules in Assymetric unit and an octamer is generated according to the symmetry. But after repeated rounds of rigid body refinement with REFMAC5 and model building with coot the R factor had been struck at 40% and R free at 50%. So my first question is whether my solution after phaser is correct. And if it is how can i lower the R factor and Rfree. The second question is it possible to have such a large difference between R factor and R free. Thanks in advance
Re: [ccp4bb] R factor R free struck
hi remember to reindex your data to P21212 in case you used Phaser to search all alternative orthorhombic SG's and it found P22121 Preben On 03/10/2010, at 04.56, Jack Russel wrote: Hi all, I have collected a data at 2.9 Å and the solved the structure using phaser . the space group comes to be P2 21 21. There are 4 molecules in Assymetric unit and an octamer is generated according to the symmetry. But after repeated rounds of rigid body refinement with REFMAC5 and model building with coot the R factor had been struck at 40% and R free at 50%. So my first question is whether my solution after phaser is correct. And if it is how can i lower the R factor and Rfree. The second question is it possible to have such a large difference between R factor and R free. Thanks in advance
[ccp4bb] R factor R free struck
Hi all, I have collected a data at 2.9 Å and the solved the structure using phaser . the space group comes to be P2 21 21. There are 4 molecules in Assymetric unit and an octamer is generated according to the symmetry. But after repeated rounds of rigid body refinement with REFMAC5 and model building with coot the R factor had been struck at 40% and R free at 50%. So my first question is whether my solution after phaser is correct. And if it is how can i lower the R factor and Rfree. The second question is it possible to have such a large difference between R factor and R free. Thanks in advance