Dear Randy, Thank you so much for your suggestion. Phenix.xtriage indicated that there are 3 possible twin operators, (-h, -k, l; h, -h-k, -l; -k, -h, -l). When I provided twin law=-h,-k,l to phenix.refine (as it seemed that only one operator can be provided), I got lower R-factors than before. Now the question is that: should I provide all the three twin operators at the same time? And how? Thanks! Best, Shun On Mar 31, 2015, at 3:26 AM, Randy Read wrote:

Dear Shun,

Phaser does a test based on the moments of the intensity distribution, after correcting for anisotropy and (if present) translational non-crystallographic symmetry. However, once a test like that has indicated that twinning is probably present, you will get a better result from running a program like phenix.xtriage, which will compare reflections related by possible twin operators and give a more precise idea of the twin fraction.

Since you’ve managed to get reasonable R-factors (even if they are higher than expected for 1.7A), the twin fraction is probably not too high. The best thing to do now is probably to run phenix.xtriage to get a suggestion for what the twin operator is, then you can provide that twin operator to phenix.refine, which will then: a) refine the twin fraction to give a much more precise estimate; b) correct for twinning in the refinement. Detwinning is not recommended any more, because it is better to refine against the original data.

Best wishes,

Randy Read

----- Randy J. Read Department of Haematology, University of Cambridge Cambridge Institute for Medical Research Tel: +44 1223 336500 Wellcome Trust/MRC Building Fax: +44 1223 336827 Hills Road E-mail: [email protected] Cambridge CB2 0XY, U.K. www-structmed.cimr.cam.ac.uk

On 31 Mar 2015, at 06:18, Shun Liu wrote:

...

Dear all, Recently I have got a data set that diffracts to 1.7 angstrom. Images look good. During data processing I did not find anything that looks strange. However, when I was doing Phaser-MR, I got a warning: “ Intensity moments suggest significant twinning (>5%). Tests based on possible twin laws will be more definitive.” What does this mean? A twinning data set? I still found a solution and refined the model step by step, using the data (20-1.7 angstrom). It seems that the final model and map are acceptable. But R-factors are 0.24/0.28, very high. Does the twin cause the high R-factors? Is there a solution to detwin? Or are the R-factors acceptable for a twinning data set? Any suggestions are appreciated and thanks in advance!

Best,

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

By the way, if you don't already know, Phenix does not support multiple twin operators. Refmac in CCP4 does, and everything is automatic. JPK -----Original Message----- From: [email protected] [mailto:[email protected]] On Behalf Of Tim Gruene Sent: Tuesday, March 31, 2015 4:49 PM To: [email protected] Subject: Re: [phenixbb] twinning Dear Shun, as you provide a twin law you basically half the data to parameter ratio, your R-factor usually drops irrespective of whether your model improved or not. You need to judge by yourself, e.g. by the quality of the map, whether or not the twin operator actually caused an improvement of your model. Best, Tim On 03/31/2015 06:47 PM, Shun Liu wrote:

Dear Randy, Thank you so much for your suggestion. Phenix.xtriage indicated that there are 3 possible twin operators, (-h, -k, l; h, -h-k, -l; -k, -h, -l). When I provided twin law=-h,-k,l to phenix.refine (as it seemed that only one operator can be provided), I got lower R-factors than before. Now the question is that: should I provide all the three twin operators at the same time? And how? Thanks!

Best, Shun

On Mar 31, 2015, at 3:26 AM, Randy Read wrote:

...

Dear Shun,

Phaser does a test based on the moments of the intensity distribution, after correcting for anisotropy and (if present) translational non-crystallographic symmetry. However, once a test like that has indicated that twinning is probably present, you will get a better result from running a program like phenix.xtriage, which will compare reflections related by possible twin operators and give a more precise idea of the twin fraction.

Since you've managed to get reasonable R-factors (even if they are higher than expected for 1.7A), the twin fraction is probably not too high. The best thing to do now is probably to run phenix.xtriage to get a suggestion for what the twin operator is, then you can provide that twin operator to phenix.refine, which will then: a) refine the twin fraction to give a much more precise estimate; b) correct for twinning in the refinement. Detwinning is not recommended any more, because it is better to refine against the original data.

Best wishes,

Randy Read

----- Randy J. Read Department of Haematology, University of Cambridge Cambridge Institute for Medical Research Tel: +44 1223 336500 Wellcome Trust/MRC Building Fax: +44 1223 336827 Hills Road E-mail: [email protected] Cambridge CB2 0XY, U.K. www-structmed.cimr.cam.ac.uk

On 31 Mar 2015, at 06:18, Shun Liu wrote:

...

Dear all, Recently I have got a data set that diffracts to 1.7 angstrom. Images look good. During data processing I did not find anything that looks strange. However, when I was doing Phaser-MR, I got a warning: " Intensity moments suggest significant twinning (>5%). Tests based on possible twin laws will be more definitive." What does this mean? A twinning data set? I still found a solution and refined the model step by step, using the data (20-1.7 angstrom). It seems that the final model and map are acceptable. But R-factors are 0.24/0.28, very high. Does the twin cause the high R-factors? Is there a solution to detwin? Or are the R-factors acceptable for a twinning data set? Any suggestions are appreciated and thanks in advance!

Best,

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

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

-- Dr Tim Gruene Institut fuer anorganische Chemie Tammannstr. 4 D-37077 Goettingen GPG Key ID = A46BEE1A

Dear Shun, There's one thing I'd like to clarify. Xtriage will list possible twin operators (based on the symmetry of the crystal lattice), and then it will generate statistics to indicate what twin fraction is obtained when you test each of those possible twin operators. So were these just the possible twin operators, or did all three of them give indications of significant twin fractions? It's often the case that there is a possible twin operator, but the crystal isn't in fact twinned using that operator. If you do indeed have multiple twin operators then, as Jacob mentioned, your only option at the moment would be Refmac5. Best wishes, Randy

On 31 Mar 2015, at 17:47, Shun Liu wrote:

Dear Randy, Thank you so much for your suggestion. Phenix.xtriage indicated that there are 3 possible twin operators, (-h, -k, l; h, -h-k, -l; -k, -h, -l). When I provided twin law=-h,-k,l to phenix.refine (as it seemed that only one operator can be provided), I got lower R-factors than before. Now the question is that: should I provide all the three twin operators at the same time? And how? Thanks!

Best, Shun

On Mar 31, 2015, at 3:26 AM, Randy Read wrote:

...

Dear Shun,

Phaser does a test based on the moments of the intensity distribution, after correcting for anisotropy and (if present) translational non-crystallographic symmetry. However, once a test like that has indicated that twinning is probably present, you will get a better result from running a program like phenix.xtriage, which will compare reflections related by possible twin operators and give a more precise idea of the twin fraction.

Since you’ve managed to get reasonable R-factors (even if they are higher than expected for 1.7A), the twin fraction is probably not too high. The best thing to do now is probably to run phenix.xtriage to get a suggestion for what the twin operator is, then you can provide that twin operator to phenix.refine, which will then: a) refine the twin fraction to give a much more precise estimate; b) correct for twinning in the refinement. Detwinning is not recommended any more, because it is better to refine against the original data.

Best wishes,

Randy Read

----- Randy J. Read Department of Haematology, University of Cambridge Cambridge Institute for Medical Research Tel: +44 1223 336500 Wellcome Trust/MRC Building Fax: +44 1223 336827 Hills Road E-mail: [email protected] Cambridge CB2 0XY, U.K. www-structmed.cimr.cam.ac.uk

On 31 Mar 2015, at 06:18, Shun Liu wrote:

...

Dear all, Recently I have got a data set that diffracts to 1.7 angstrom. Images look good. During data processing I did not find anything that looks strange. However, when I was doing Phaser-MR, I got a warning: “ Intensity moments suggest significant twinning (>5%). Tests based on possible twin laws will be more definitive.” What does this mean? A twinning data set? I still found a solution and refined the model step by step, using the data (20-1.7 angstrom). It seems that the final model and map are acceptable. But R-factors are 0.24/0.28, very high. Does the twin cause the high R-factors? Is there a solution to detwin? Or are the R-factors acceptable for a twinning data set? Any suggestions are appreciated and thanks in advance!

Best,

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

------ Randy J. Read Department of Haematology, University of Cambridge Cambridge Institute for Medical Research Tel: + 44 1223 336500 Wellcome Trust/MRC Building Fax: + 44 1223 336827 Hills Road E-mail: [email protected] Cambridge CB2 0XY, U.K. www-structmed.cimr.cam.ac.uk

Dear Randy, Below is the statistics depending on twin laws by Xtriage: OPERATOR TYPE R OBS. BRITTON ALPHA H ALPHA ML ALPHA -h,-k,l M 0.395 0.089 0.088 0.022 h,-h-k,-l M 0.059 0.440 0.447 0.435 -k,-h,-l M 0.404 0.082 0.087 0.022 The second one (h,-h-k,-l) gives a significant twin fraction, indicating an actual twin law (Am l right?). How about the other two? Low twin fraction or no twin? What is the criteria? Which operators should I choose for refinement? Only the second one? Thanks! Best, Shun On Apr 1, 2015, at 12:38 AM, Randy Read wrote:

Dear Shun,

There's one thing I'd like to clarify. Xtriage will list possible twin operators (based on the symmetry of the crystal lattice), and then it will generate statistics to indicate what twin fraction is obtained when you test each of those possible twin operators. So were these just the possible twin operators, or did all three of them give indications of significant twin fractions? It's often the case that there is a possible twin operator, but the crystal isn't in fact twinned using that operator.

If you do indeed have multiple twin operators then, as Jacob mentioned, your only option at the moment would be Refmac5.

Best wishes,

Randy

...

On 31 Mar 2015, at 17:47, Shun Liu wrote:

Dear Randy, Thank you so much for your suggestion. Phenix.xtriage indicated that there are 3 possible twin operators, (-h, -k, l; h, -h-k, -l; -k, -h, -l). When I provided twin law=-h,-k,l to phenix.refine (as it seemed that only one operator can be provided), I got lower R-factors than before. Now the question is that: should I provide all the three twin operators at the same time? And how? Thanks!

Best, Shun

On Mar 31, 2015, at 3:26 AM, Randy Read wrote:

...

Dear Shun,

Phaser does a test based on the moments of the intensity distribution, after correcting for anisotropy and (if present) translational non-crystallographic symmetry. However, once a test like that has indicated that twinning is probably present, you will get a better result from running a program like phenix.xtriage, which will compare reflections related by possible twin operators and give a more precise idea of the twin fraction.

Since you’ve managed to get reasonable R-factors (even if they are higher than expected for 1.7A), the twin fraction is probably not too high. The best thing to do now is probably to run phenix.xtriage to get a suggestion for what the twin operator is, then you can provide that twin operator to phenix.refine, which will then: a) refine the twin fraction to give a much more precise estimate; b) correct for twinning in the refinement. Detwinning is not recommended any more, because it is better to refine against the original data.

Best wishes,

Randy Read

----- Randy J. Read Department of Haematology, University of Cambridge Cambridge Institute for Medical Research Tel: +44 1223 336500 Wellcome Trust/MRC Building Fax: +44 1223 336827 Hills Road E-mail: [email protected] Cambridge CB2 0XY, U.K. www-structmed.cimr.cam.ac.uk

On 31 Mar 2015, at 06:18, Shun Liu wrote:

...

Dear all, Recently I have got a data set that diffracts to 1.7 angstrom. Images look good. During data processing I did not find anything that looks strange. However, when I was doing Phaser-MR, I got a warning: “ Intensity moments suggest significant twinning (>5%). Tests based on possible twin laws will be more definitive.” What does this mean? A twinning data set? I still found a solution and refined the model step by step, using the data (20-1.7 angstrom). It seems that the final model and map are acceptable. But R-factors are 0.24/0.28, very high. Does the twin cause the high R-factors? Is there a solution to detwin? Or are the R-factors acceptable for a twinning data set? Any suggestions are appreciated and thanks in advance!

Best,

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

------ Randy J. Read Department of Haematology, University of Cambridge Cambridge Institute for Medical Research Tel: + 44 1223 336500 Wellcome Trust/MRC Building Fax: + 44 1223 336827 Hills Road E-mail: [email protected] Cambridge CB2 0XY, U.K. www-structmed.cimr.cam.ac.uk

Dear Randy, Peter, Jacob, and Tim, Thank you all very much for your instructive suggestions and comments. Best, Shun On Apr 1, 2015, at 8:48 AM, Randy Read wrote:

Hi,

It looks like only the second one indicates significant twinning. For that one, the R-factor between twin-related reflections is very low while the estimated twinning fraction (by all three tests) is around 0.44, which is close enough to 0.5 that this might be a perfectly twinned crystal. So you only want to specify the second twinning operator to phenix.refine, and you don't need to worry about the other two potential operators. For these the R-factor between twin-related reflections is much larger and the estimated twin fractions are much lower. It's possible there's some pseudosymmetry, which could give R-factors lower than random and estimated twin fractions a bit larger than 0.

Regards,

Randy

...

On 1 Apr 2015, at 16:33, Shun Liu wrote:

Dear Randy, Below is the statistics depending on twin laws by Xtriage:

OPERATOR TYPE R OBS. BRITTON ALPHA H ALPHA ML ALPHA -h,-k,l M 0.395 0.089 0.088 0.022 h,-h-k,-l M 0.059 0.440 0.447 0.435 -k,-h,-l M 0.404 0.082 0.087 0.022

The second one (h,-h-k,-l) gives a significant twin fraction, indicating an actual twin law (Am l right?). How about the other two? Low twin fraction or no twin? What is the criteria? Which operators should I choose for refinement? Only the second one? Thanks!

Best,

Shun

On Apr 1, 2015, at 12:38 AM, Randy Read wrote:

...

Dear Shun,

There's one thing I'd like to clarify. Xtriage will list possible twin operators (based on the symmetry of the crystal lattice), and then it will generate statistics to indicate what twin fraction is obtained when you test each of those possible twin operators. So were these just the possible twin operators, or did all three of them give indications of significant twin fractions? It's often the case that there is a possible twin operator, but the crystal isn't in fact twinned using that operator.

If you do indeed have multiple twin operators then, as Jacob mentioned, your only option at the moment would be Refmac5.

Best wishes,

Randy

...

On 31 Mar 2015, at 17:47, Shun Liu wrote:

Dear Randy, Thank you so much for your suggestion. Phenix.xtriage indicated that there are 3 possible twin operators, (-h, -k, l; h, -h-k, -l; -k, -h, -l). When I provided twin law=-h,-k,l to phenix.refine (as it seemed that only one operator can be provided), I got lower R-factors than before. Now the question is that: should I provide all the three twin operators at the same time? And how? Thanks!

Best, Shun

On Mar 31, 2015, at 3:26 AM, Randy Read wrote:

...

Dear Shun,

Phaser does a test based on the moments of the intensity distribution, after correcting for anisotropy and (if present) translational non-crystallographic symmetry. However, once a test like that has indicated that twinning is probably present, you will get a better result from running a program like phenix.xtriage, which will compare reflections related by possible twin operators and give a more precise idea of the twin fraction.

Since you’ve managed to get reasonable R-factors (even if they are higher than expected for 1.7A), the twin fraction is probably not too high. The best thing to do now is probably to run phenix.xtriage to get a suggestion for what the twin operator is, then you can provide that twin operator to phenix.refine, which will then: a) refine the twin fraction to give a much more precise estimate; b) correct for twinning in the refinement. Detwinning is not recommended any more, because it is better to refine against the original data.

Best wishes,

Randy Read

----- Randy J. Read Department of Haematology, University of Cambridge Cambridge Institute for Medical Research Tel: +44 1223 336500 Wellcome Trust/MRC Building Fax: +44 1223 336827 Hills Road E-mail: [email protected] Cambridge CB2 0XY, U.K. www-structmed.cimr.cam.ac.uk

On 31 Mar 2015, at 06:18, Shun Liu wrote:

...

Dear all, Recently I have got a data set that diffracts to 1.7 angstrom. Images look good. During data processing I did not find anything that looks strange. However, when I was doing Phaser-MR, I got a warning: “ Intensity moments suggest significant twinning (>5%). Tests based on possible twin laws will be more definitive.” What does this mean? A twinning data set? I still found a solution and refined the model step by step, using the data (20-1.7 angstrom). It seems that the final model and map are acceptable. But R-factors are 0.24/0.28, very high. Does the twin cause the high R-factors? Is there a solution to detwin? Or are the R-factors acceptable for a twinning data set? Any suggestions are appreciated and thanks in advance!

Best,

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

------ Randy J. Read Department of Haematology, University of Cambridge Cambridge Institute for Medical Research Tel: + 44 1223 336500 Wellcome Trust/MRC Building Fax: + 44 1223 336827 Hills Road E-mail: [email protected] Cambridge CB2 0XY, U.K. www-structmed.cimr.cam.ac.uk

------ Randy J. Read Department of Haematology, University of Cambridge Cambridge Institute for Medical Research Tel: + 44 1223 336500 Wellcome Trust/MRC Building Fax: + 44 1223 336827 Hills Road E-mail: [email protected] Cambridge CB2 0XY, U.K. www-structmed.cimr.cam.ac.uk