TaBr SAD Autosol and Phaser

older
linking only one conformer to a...

Ian S

4 Apr 2018 4 Apr '18

11:22 p.m.

Hi Phenixbb, I trying to phase a structure soaked in TaBr and am running into some technical problems. My anomalous signal goes out to about 4.5 and I have a 1.9A native data set. SIR hasn't worked so far. Autosol SAD has given me several convincing solutions when searching for Tantalum (not TX), finding 11 sites (not clustered) at 6 angstroms: FOM =0.478, map=0.1.3, and corr of 0.82 and a final score of 37+-15.71. By only playing around with the sites I can get an overall score of 73 +- 4.58, map skew to 0.5, and a RMS density to 0.79, thought the R-factor increases to .43 and the FOM . Perhaps not suprisingly at this resolution, the map is uninterpretable (maybe some helices) so I'm trying to improve the phases by model building and by using Phaser and giving it TX sites. If I put the Ta sites in Phaser with the same dataset as Autosol, I get a FOM of .342. If I use the TX label it drops to 0.143. What is happening here? I know that typically TaBr has one or two major sites (my protein isnt massive, only about 1200 aa predicted in the ASU) , so I use only the top two major sites and have the same problem. Does this indicate that the Ta sites Autosol found are not in fact the sites I'm looking for? Or maybe that there really are 10 TaBr sites all with low occupancy? Any suggestions would be appreciated. Thank you. Best, Ian

Attachments:

attachment.html (text/html — 1.5 KB)

Show replies by date

Randy Read

5 Apr 5 Apr

8:52 a.m.

Dear Ian, If you could share the data (off-list, of course), I would be interested in looking to see why the cluster SAD phasing is giving results that are so much poorer than for individual Ta atoms. It’s possible that something isn’t being done optimally — this feature in Phaser hasn’t been exercised as much as other features. If SIR hasn’t worked, that might well imply that the native and TaBr crystals aren’t completely isomorphous. In that case, cross-crystal averaging could be very useful. If you can approximately define the volume covering the protein in the TaBr-phased map, you can cut that density out and use it to phase the native crystal by molecular replacement. (Actually, just rigid-body refinement usually works and gives you operators near the identity, which makes life easier.) Depending on how different the crystal forms are, cross-crystal averaging of the two maps might yield enough improvement to either start to see the individual atoms in the clusters or to place helices more accurately. Once you have a map from averaging or a model that has higher resolution features, log-likelihood-gradient completion in Phaser is more likely to distinguish the individual atoms, which should help to push the resolution for which you get good phases. I’m not sure why Phaser isn’t reproducing the FOM values you got in AutoSol. If you have only provided the SAD data to AutoSol, then it should be using Phaser internally. Have you specified different resolution limits in the two approaches? 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 5 Apr 2018, at 00:22, Ian S wrote:

Hi Phenixbb,

I trying to phase a structure soaked in TaBr and am running into some technical problems. My anomalous signal goes out to about 4.5 and I have a 1.9A native data set. SIR hasn't worked so far.

Autosol SAD has given me several convincing solutions when searching for Tantalum (not TX), finding 11 sites (not clustered) at 6 angstroms: FOM =0.478, map=0.1.3, and corr of 0.82 and a final score of 37+-15.71. By only playing around with the sites I can get an overall score of 73 +- 4.58, map skew to 0.5, and a RMS density to 0.79, thought the R-factor increases to .43 and the FOM .

Perhaps not suprisingly at this resolution, the map is uninterpretable (maybe some helices) so I'm trying to improve the phases by model building and by using Phaser and giving it TX sites. If I put the Ta sites in Phaser with the same dataset as Autosol, I get a FOM of .342. If I use the TX label it drops to 0.143. What is happening here? I know that typically TaBr has one or two major sites (my protein isnt massive, only about 1200 aa predicted in the ASU) , so I use only the top two major sites and have the same problem. Does this indicate that the Ta sites Autosol found are not in fact the sites I'm looking for? Or maybe that there really are 10 TaBr sites all with low occupancy? Any suggestions would be appreciated. Thank you.

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

Edward A. Berry

5:41 p.m.

New subject: rigid body refinement with electron density as a model

Is it possible to tell phaser to just do rigid body refinement? Or how do you do rigid body with electron density? On 04/05/2018 04:52 AM, Randy Read wrote:

...

you can cut that density out and use it to phase the native crystal by molecular replacement. (Actually, just rigid-body refinement usually works

Randy Read

7:12 p.m.

New subject: rigid body refinement with electron density as a model

Hi, In the Phaser-MR GUI in Phenix, this is relatively easy for a PDB model, though I see it’s not set up properly for a map ensemble! You choose “Refinement and phasing” for the Phaser mode. You define a PDB file for the model in the right place as the ensemble and check the box saying “Ensemble is fixed partial solution”. You still have to define the composition, but for this mode you don’t specify a search procedure because there’s no search taking place. In principle it should work basically the same way for a map ensemble, but I guess I’ve always done that from a script file! Now I see there isn’t a check box for “fixed partial solution” when the model is defined with a map. So you have to set up a script to run Phaser, in the same way as for a coordinate file, but you use the ENSEMBLE <MODLID> HKLIN command to define the model through map coefficients. One trick here is that when you cut out the density, you have to make sure not to shift it away from its current position to the origin — this is one of the things controlled by “for_phaser=True” in phenix.cut_out_density. Then you can say that it’s already in about the right position with the command “SOLU ORIGIN ENSEMBLE <MODLID>”. Let me know if you need any more help? In the meantime, we should change the GUI so that this option can be selected with a check box… Randy ----- 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 5 Apr 2018, at 18:41, Edward A. Berry wrote:

Is it possible to tell phaser to just do rigid body refinement? Or how do you do rigid body with electron density?

On 04/05/2018 04:52 AM, Randy Read wrote:

...
you can cut that density out and use it to phase the native crystal by molecular replacement. (Actually, just rigid-body refinement usually works

Edward A. Berry

8 p.m.

New subject: rigid body refinement with electron density as a model

I'm mostly using .eff files, and I do seem to have a line defining that parameter: solution_origin_ensemble = None So I would replace "None" with the ID of the ensemble if it is nearly in the right place? Like solving a structure using the structure from an almost-isomorphous crystal. I'm asking mainly for future reference, but I will use this feature with atomic models to get the same AU for different crystals. Would it have a larger radius of convergence than conventional rigid-body refinement? On 04/05/2018 03:12 PM, Randy Read wrote:

...

Hi,

In the Phaser-MR GUI in Phenix, this is relatively easy for a PDB model, though I see it’s not set up properly for a map ensemble! You choose “Refinement and phasing” for the Phaser mode. You define a PDB file for the model in the right place as the ensemble and check the box saying “Ensemble is fixed partial solution”. You still have to define the composition, but for this mode you don’t specify a search procedure because there’s no search taking place.

In principle it should work basically the same way for a map ensemble, but I guess I’ve always done that from a script file! Now I see there isn’t a check box for “fixed partial solution” when the model is defined with a map. So you have to set up a script to run Phaser, in the same way as for a coordinate file, but you use the ENSEMBLE <MODLID> HKLIN command to define the model through map coefficients. One trick here is that when you cut out the density, you have to make sure not to shift it away from its current position to the origin — this is one of the things controlled by “for_phaser=True” in phenix.cut_out_density. Then you can say that it’s already in about the right position with the command “SOLU ORIGIN ENSEMBLE <MODLID>”.

Let me know if you need any more help? In the meantime, we should change the GUI so that this option can be selected with a check box…

Randy

----- 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 5 Apr 2018, at 18:41, Edward A. Berry wrote:

Is it possible to tell phaser to just do rigid body refinement? Or how do you do rigid body with electron density?

On 04/05/2018 04:52 AM, Randy Read wrote:

...
you can cut that density out and use it to phase the native crystal by molecular replacement. (Actually, just rigid-body refinement usually works

Randy Read

6 Apr 6 Apr

9:26 a.m.

New subject: rigid body refinement with electron density as a model

Hi, I had to run some test jobs to check how this would work in a .eff file. So, actually, what you need to do is to go to the ensemble section of the .eff file where the model to be fixed is defined, and set "solution_at_origin = True". That works for both models specified as PDB files or as density, as long as they're close to being in the correct position. Our experience with the rigid-body refinement in Phaser has been very good in terms of radius of convergence, though we haven't done systematic tests against phenix.refine and Refmac to see how they would compare. Definitely it's a good option to try first if you have a crystal that should be reasonably isomorphous and it will usually work in those circumstances. If it doesn't work, then you can go back to trying a full MR search. Even in that case (at least if you're running Phaser from a script) you can use the SOLU TEMPLATE command to indicate the start of a target solution that you would like the answer to match to as closely as possible, which should work even if there are several components being allowed to move. However, I don't see a way at the moment of specifying that from phil .eff files! Randy

...

On 5 Apr 2018, at 21:00, Edward A. Berry wrote:

I'm mostly using .eff files, and I do seem to have a line defining that parameter: solution_origin_ensemble = None So I would replace "None" with the ID of the ensemble if it is nearly in the right place? Like solving a structure using the structure from an almost-isomorphous crystal.

I'm asking mainly for future reference, but I will use this feature with atomic models to get the same AU for different crystals. Would it have a larger radius of convergence than conventional rigid-body refinement?

On 04/05/2018 03:12 PM, Randy Read wrote:

...
Hi,

In the Phaser-MR GUI in Phenix, this is relatively easy for a PDB model, though I see it’s not set up properly for a map ensemble! You choose “Refinement and phasing” for the Phaser mode. You define a PDB file for the model in the right place as the ensemble and check the box saying “Ensemble is fixed partial solution”. You still have to define the composition, but for this mode you don’t specify a search procedure because there’s no search taking place.

In principle it should work basically the same way for a map ensemble, but I guess I’ve always done that from a script file! Now I see there isn’t a check box for “fixed partial solution” when the model is defined with a map. So you have to set up a script to run Phaser, in the same way as for a coordinate file, but you use the ENSEMBLE <MODLID> HKLIN command to define the model through map coefficients. One trick here is that when you cut out the density, you have to make sure not to shift it away from its current position to the origin — this is one of the things controlled by “for_phaser=True” in phenix.cut_out_density. Then you can say that it’s already in about the right position with the command “SOLU ORIGIN ENSEMBLE <MODLID>”.

Let me know if you need any more help? In the meantime, we should change the GUI so that this option can be selected with a check box…

Randy

----- 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 5 Apr 2018, at 18:41, Edward A. Berry wrote:

Is it possible to tell phaser to just do rigid body refinement? Or how do you do rigid body with electron density?

On 04/05/2018 04:52 AM, Randy Read wrote:

...
you can cut that density out and use it to phase the native crystal by molecular replacement. (Actually, just rigid-body refinement usually works

------ 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

Edward A. Berry

9 Apr 9 Apr

3:11 p.m.

New subject: rigid body refinement with electron density as a model

On 04/05/2018 04:00 PM, Edward A. Berry wrote:

...

I'm mostly using .eff files, and I do seem to have a line defining that parameter: solution_origin_ensemble = None So I would replace "None" with the ID of the ensemble if it is nearly in the right place?

In case anyone else is running phaser from .eff files, No that is NOT the way. Leave that as "None" or comment it out and use: phaser { mode = ANO CCA EP_AUTO MR_AUTO MR_FRF MR_FTF MR_PAK *MR_RNP NMAXYZ SCEDS and: ensemble { solution_at_origin = true }} On 04/06/2018 05:26 AM, Randy Read wrote:

...

I had to run some test jobs to check how this would work in a .eff file. So, actually, what you need to do is to go to the ensemble section of the .eff file where the model to be fixed is defined, and set "solution_at_origin = True". That works for both models specified as PDB files or as density, as long as they're close to being in the correct position.

2455

Age (days ago)

2460

Last active (days ago)

List overview

Download

6 comments

3 participants

participants (3)

Edward A. Berry
Ian S
Randy Read