[phenixbb] MR-SAD, SAD

Mon Jan 26 02:14:42 PST 2015

Dear Almudena,

If AutoSol is using the Hendrickson-Lattman coefficients from the SAD phasing, then both the refinement and the map calculation should be using that phase information, so running MR-SAD shouldn’t necessarily improve things.  However, what can happen is that, with the addition of a partial protein model, the MR-SAD procedure finds a more complete and more accurate substructure, which then can yield significantly better phases and maps.  We have a nice example of this from a test case (in the appendix of this paper: http://journals.iucr.org/d/issues/2010/04/00/ba5142/index.html), where a substructure containing 57 atoms allowed a model of 70% of nitrate reductase to be built.  MR-SAD with this model found a substructure containing 105 atoms, and the next step of model-building built a substantially more complete model.  In this case, there are 21 Fe atoms (mostly in Fe-S clusters), one Mo atom, and over 120 S or P atoms, so there are a lot of minor sites to find.

As Tom mentioned, Phaser produces two types of Hendrickson Lattman coefficients, the “HL” ones that contain all phase information (including that from the partial structure) and the “HLanom” ones that contain only the phase information from the anomalous scattering.  The HL variant should be used when you need the complete phase information, such as for density modification.  However, if you are refining a model that includes the atoms in the partial structure used for MR-SAD, you want to use the HLanom coefficients to avoid including the partial structure phase information twice.

An average FOM of 0.49 is very good for SAD phasing, because the 2-fold phase ambiguity tends to limit phase accuracy.  We’re looking into how to interpret the absolute LLG values, but at the moment I can’t tell you how good that number is (though it looks fairly high, which is probably good)!

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: rjr27 at cam.ac.uk
Cambridge CB2 0XY, U.K.                               www-structmed.cimr.cam.ac.uk

On 26 Jan 2015, at 08:40, Almudena Ponce Salvatierra <maps.farma at GMAIL.COM> wrote:

> HI!
> 
> George gave me an idea actually... I did SAD phasing with Co sites and with AutoSol I got a partial model, that I then edited. However, since you were suggesting the phases should be better when you combine those from the model with the experimental ones I ran Phaser-EP in SAD-MR mode. I have provided the model I edited after AutoSol, my HA sites, and the data, and I got a better map than even just with AutoSol!!! I mean, when I submit it to Resolve, after phaser. :-) 
> 
> Is it a good solution if the FOM is 0.49 and LLG is 36449.63 after Phaser? 
> 
> Sooo thanks a lot for the inspiration! (it's not so easy on Monday mornings!!) 
> 
> Best wishes, 
> 
> Almudena. 
> 
> 2015-01-25 21:06 GMT+01:00 Terwilliger, Thomas Charles <terwilliger at lanl.gov>:
> Hi Georg,
> 
> I would say probably neither a bug nor exactly "normal"...
> 
> If everything were ideal, then an MR solution, even if very distant and contributing very little phase information, should improve the Phaser identification of the sub-structure.  Therefore as you imagine, including it should have improved the situation.  In real life there is a lot of noise in the system, so small changes can sometimes make the difference between finding sites and not. Also the default parameters for Phaser sub-structure identification are not exactly the same for MR-SAD and for extreme defaults in autosol.
> 
> I am guessing that autosol, with extreme defaults, managed to find something that was partially correct, then the iteration of the substructure search (using density from the density-modified map and/or partial model that was built) resulted in additional sites, while MR-SAD did not happen to identify any convincing sites and autosol stopped.
> 
> On your second question...the HL coefficients from MR-SAD are actually produced twice in the output MTZ file; once with information from the model  (HLA HLB etc) and once without information from the model. (HLanomA HLanomB etc). I think in general your intuition is right and the best map will come from including model information.  However if you would like to reduce model bias, you might want to exclude model information and use the HLanomA etc.  The map coefficients (FWT PHWT) in the overall_best_denmod_map_coeffs.mtz from MRSAD will include or not include the model information based on how you set the parameter use_hl_anom_in_denmod (default is False, use HL coeffs and include model information).
> 
> I'll pass on the other requests!
> 
> All the best,
> Tom T
> 
> 
> ________________________________________
> From: phenixbb-bounces at phenix-online.org [phenixbb-bounces at phenix-online.org] on behalf of Georg Mlynek [georg.mlynek at univie.ac.at]
> Sent: Sunday, January 25, 2015 4:54 AM
> To: phenixbb at phenix-online.org
> Subject: [phenixbb] MR-SAD, SAD
> 
> Dear phenix-developers,
> 
> I have collected a high redundancy dataset on our bruker home-source to
> get good anomalous data to confirm if a ligandbound near the active site
> is really the one I think.
> 
> 1. If I do MR with phaser-MR a solution is found easiliy. If I then
> continue with MR-SAD the program does not give me a solution.
> 
> No file named overall_best.pdb is  present in the output directory.
> 
> Warnings: FOM < 0.05 ... skipping solution 1
> 
> However Autosol with extreme density modification and  Iterate
> substructure search is able to solve the structure.
> 
> Is this a bug or normal.
> 
> 2. A follow up on this: Will be the MTZ file (phases) produced from
> MR-SAD always better (compared to SAD or MR alone) because it will
> contain phases that combine the information from both the MR model and
> the SAD data. Or can there be cases where bad phases form either MR or
> SAD will mess up the other one.
> 
> 3. There is a small bug in merging statistics. cc_ano is just writen out
> in the log output tab but not in the summary tab.
> 
> 4. It would be nice, if xtriage could also print how many % of the
> reflections are Rfree flagged.
> 
> Thanks in advance, best regards, Georg.
> _______________________________________________
> phenixbb mailing list
> phenixbb at phenix-online.org
> http://phenix-online.org/mailman/listinfo/phenixbb
> 
> _______________________________________________
> phenixbb mailing list
> phenixbb at phenix-online.org
> http://phenix-online.org/mailman/listinfo/phenixbb
> 
> 
> 
> -- 
> Almudena Ponce-Salvatierra
> Macromolecular crystallography and Nucleic acid chemistry
> Max Planck Institute for Biophysical Chemistry
> Am Fassberg 11 37077 Göttingen
> Germany
> 