Hi Esko, thanks for the paper - very interesting! Perhaps accounting for ptNCS matters a lot for structure solution or/and refinement/model building at early stages, however I did not see any problem with refinement in this particular example. Here is what I did: 1) Get model and data from PDB: phenix.fetch_pdb 1yup --mtz 2) Get initial statistics: R-factors from PDB file header, geometry using command: phenix.pdbtools model_stat=true 1yup.pdb Rwork=0.235 Rfree=0.311 all-atom clashscore : 8.21 ramachandran plot: outliers : 1.52 % allowed : 7.43 % favored : 91.05 % rotamer outliers : 14.11 % cbeta deviations : 3 3) Then I perturbed the model and refined it with phenix.refine (restrained coordinates and B-factors, water update, weights optimization, local real-space refinement). Here is the statistics for refined model: Rwork=0.2380 Rfree=0.2709 all-atom clashscore : 10.61 ramachandran plot: outliers : 0.80 % allowed : 5.20 % favored : 94.00 % rotamer outliers : 2.38 % cbeta deviations : 0 Just to make it clear: I'm *not* saying we shouldn't account for ptNCS in refinement; in fact, we really should especially given that Randy Read et al worked out all the methods for this! I will re-visit this once we implement ptNCS in phenix.refine. All the best, Pavel On 4/3/15 9:01 AM, Esko Oksanen wrote:

Hi Mark,

I believe at least part of the problem arises already in the scaling stage, where the programs assume a unimodal distribution which is not the case in the presence of pseudo-translations. We encountered a similar issue with refinement and pseudo-translation, where I simply scaled the systematically weak data separately from the strong and did rigid body refinement against the weak data and restrained against the strong. The details can be found in http://scripts.iucr.org/cgi-bin/paper?ea5052.

Best, Esko *_____________________* * * *Esko Oksanen, PhD* Instrument Scientist - Neutron Macromolecular Crystallography

European Spallation Source ESS AB P.O Box 176, SE-221 00 Lund, Sweden Visiting address: Tunavägen 24, Lund

Switchboard: +46 46 888 00 00 Phone: +46 46 888 30 36 Mobile: +46 721 79 20 36 E-mail: [email protected] mailto:[email protected]

www.esss.se http://www.esss.se/

On 03 Apr 2015, at 15:59, Mark A Saper mailto:[email protected]> wrote:

...

I know that Phaser takes account of NCS translation in the likelihood statistics ( Read, Adams, McCoy (2013) Intensity statistics in the presence of translational non crystallographic symmetry”). Has this also been incorporated into phenix.refine? I have a structure with pseudo NCS translations because of NCS rotational symmetry. One suggestion that I found in the archives, was to explicitly LS target refinement rather than ML. Any comments?

Mark _________________________________ Mark A. Saper, Ph.D. Associate Professor of Biological Chemistry, University of Michigan Ann Arbor MI 48109-1055 U.S.A. [email protected] mailto:[email protected] phone (734) 764-3353 fax (734) 764-3323

_______________________________________________ phenixbb mailing list [email protected] mailto:[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]

My R-factors are similar to what Pavel obtained. My Patterson peak was 43% of origin. Haven't read Esko's response yet. Mark Saper 3040 & 3250 Chemistry Building Dept of Biological Chemistry [email protected] 734 764 3353 office 734 276 6505 mobile

On Apr 3, 2015, at 1:58 PM, Pavel Afonine wrote:

Hi Esko,

thanks for the paper - very interesting!

Perhaps accounting for ptNCS matters a lot for structure solution or/and refinement/model building at early stages, however I did not see any problem with refinement in this particular example.

Here is what I did:

1) Get model and data from PDB:

phenix.fetch_pdb 1yup --mtz

2) Get initial statistics: R-factors from PDB file header, geometry using command:

phenix.pdbtools model_stat=true 1yup.pdb

Rwork=0.235 Rfree=0.311 all-atom clashscore : 8.21 ramachandran plot: outliers : 1.52 % allowed : 7.43 % favored : 91.05 % rotamer outliers : 14.11 % cbeta deviations : 3

3) Then I perturbed the model and refined it with phenix.refine (restrained coordinates and B-factors, water update, weights optimization, local real-space refinement). Here is the statistics for refined model:

Rwork=0.2380 Rfree=0.2709 all-atom clashscore : 10.61 ramachandran plot: outliers : 0.80 % allowed : 5.20 % favored : 94.00 % rotamer outliers : 2.38 % cbeta deviations : 0

Just to make it clear: I'm *not* saying we shouldn't account for ptNCS in refinement; in fact, we really should especially given that Randy Read et al worked out all the methods for this!

I will re-visit this once we implement ptNCS in phenix.refine.

All the best, Pavel

...

On 4/3/15 9:01 AM, Esko Oksanen wrote: Hi Mark,

I believe at least part of the problem arises already in the scaling stage, where the programs assume a unimodal distribution which is not the case in the presence of pseudo-translations. We encountered a similar issue with refinement and pseudo-translation, where I simply scaled the systematically weak data separately from the strong and did rigid body refinement against the weak data and restrained against the strong. The details can be found in http://scripts.iucr.org/cgi-bin/paper?ea5052.

Best, Esko _____________________

Esko Oksanen, PhD Instrument Scientist - Neutron Macromolecular Crystallography

European Spallation Source ESS AB P.O Box 176, SE-221 00 Lund, Sweden Visiting address: Tunavägen 24, Lund

Switchboard: +46 46 888 00 00 Phone: +46 46 888 30 36 Mobile: +46 721 79 20 36 E-mail: [email protected]

www.esss.se

...

On 03 Apr 2015, at 15:59, Mark A Saper wrote:

I know that Phaser takes account of NCS translation in the likelihood statistics ( Read, Adams, McCoy (2013) Intensity statistics in the presence of translational non crystallographic symmetry”). Has this also been incorporated into phenix.refine? I have a structure with pseudo NCS translations because of NCS rotational symmetry. One suggestion that I found in the archives, was to explicitly LS target refinement rather than ML. Any comments?

Mark _________________________________ Mark A. Saper, Ph.D. Associate Professor of Biological Chemistry, University of Michigan Ann Arbor MI 48109-1055 U.S.A. [email protected] phone (734) 764-3353 fax (734) 764-3323

_______________________________________________ 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]