Satvik - if you attach your data processing logs I can tell you what to look out for - too abstract to do without a concrete example to discuss.
Eleanor

On 20 September 2017 at 14:00, <[log in to unmask]> wrote:

Dear Satvik,

 

An R/Rfree of 0.29/0.35 after one round of automatic model building indicates that your solution is correct. You can proceed with refinement and rebuilding. You can take either the monomer-based or dimer-based solution, it does not matter. Personally I would take the monomer-based solution since it allowed for some small difference in orientation of the monomers within the dimer.

 

Best,

Herman

 

Von: CCP4 bulletin board [mailto:CCP4BB@JISCMAIL.AC.UK] Im Auftrag von Satvik Kumar
Gesendet: Mittwoch, 20. September 2017 14:40
An: [log in to unmask]
Betreff: Re: [ccp4bb] doubt regarding MR search model

 

Hello,

Thanks all for your valuable suggestions. Two things that stuck me were "one gets to know if the space group is correct only after solving the structure" and "to try all possible space groups until one lands up with the correct solution".

I ran Phaser with "all alternative space groups". The solution is still in the same space group P212121. This holds true with both the monomer as well as dimer search models. I hope I have checked all orthorhombic space groups by doing this as I am unsure what "all alternative space groups" means.

I also ran Zanuda which tells me that the space group P212121 is correct in both cases.

Pointless output indicates that "Systematic absence probability: 0.818". Can I take this to be convincing or do I need to run other tests?

 

I dont know how to check for non-crystallographic symmetry? Please tell me how I can do so.

 

The model is 60 percent identical to my protein. Also, one round of automated model building has brought down the R values. The Rwork,Rfree stand at 0.29,0.35 (monomer search model) and 0.30, 0.35 (dimer search model). There is good scope for building residues in both cases. But which one do I go ahead with is my question.

 

Thanks,

Satvik

 

 

On Tue, Sep 19, 2017 at 11:40 PM, Eleanor Dodson <[log in to unmask]> wrote:

Well - you haven't said what the sequence identity between model and your protein is, nor if you have a non-crystallographic translation. 

 

With low homology that R factor drop is acceptable and rebuilding can fix it, However if there is high homology you might expect better.

 

But this sort of conjecture is pretty pointless - check in all orthorhombic  space groups as Mark suggests.

 

Eleanor

 

On 19 September 2017 at 15:16, Mark J van Raaij <[log in to unmask]> wrote:

With Rs of 43/48% I don't think you can be sure that your spacegroup is right.

You should always try all the spacegroup possibilities until you get a solution you are sure is right, i.e. that refines to Rs of around 35% or preferably even lower.

More so in the case of screw axes, so try P222, P2122, P2212, P2221, P21212, P21221, P22121 and P212121. Phaser can do this automatically for you by clicking the right box. 

If necessary, then try lower symmetry like P21 and perhaps P1.

Programs like Xanuda can help.

 

 

Mark J van Raaij
Dpto de Estructura de Macromoleculas
Centro Nacional de Biotecnologia - CSIC
calle Darwin 3
E-28049 Madrid, Spain
tel. (+34) 91 585 4616
http://wwwuser.cnb.csic.es/~mjvanraaij

Editor of Acta Crystallographica F, Structural Biology Communications

 

On 19 Sep 2017, at 16:01, Satvik Kumar <[log in to unmask]> wrote:

 

Hello,

Thanks everyone for your explanations.

I have pasted the pointless output to provide more information.

Best Solution:                                  space group P 21 21 21
Laue group probability:                   0.959
Systematic absence probability:     0.818
Total probability:                             0.785
Space group confidence:                0.751
Laue group confidence                   0.951

 

Unit cell:   82.10 100.51 157.11     90.00  90.00  90.00
Also based on L-test, pointless says data does not suggest twinning.
Yes, the R values go down when I refine in both cases. After 20 rounds of restrained refinement using the coordinates generated by monomer as search model, the Rwork and Rfree are 0.43 and 0.48

respectively. Refinement using the coordinates generated by using dimer as search model also results in similar R values. I have attached the plots to show that the R values indeed reduce in both cases.

Is my space group correct? Do I need to reexamine the space group even though the probability is high?

If my space group is indeed correct, how do I decide whether to go ahead with the results generated by the monomer search model or the dimer?

Please share your thoughts.

Thanks,

Satvik

 

On Mon, Sep 18, 2017 at 7:36 PM, Eleanor Dodson <[log in to unmask]> wrote:

You need to provide a bit more information.

First of all about the data processing..

Is the space group correct?

ways of being misled are:

Non-crystallographic translations with a shift of ~0.5 along an axis - say a.  This will generate absences in the odd h 0 0 reflections and can make the space group appear to be P 21 21 21 whilst it is really P 2 21 21..

Perfect twinning can have the same effect. In an orthorhombix space group this can usually only occur if two axes have approximately the same length, but the data processing stats can indicate if that is the case.

Then - re PHASER. The packing rejection criteria may be set too severely - that seems the case for your solution.

Best check on any MR solution is: does it refine - give it 20 cycles of mindless refinement and see if the R and FreeR go down.

Then look at the maps and see if there are obvious corrections to be made..

Eleanor

 

On 18 September 2017 at 14:59, Satvik Kumar <[log in to unmask]> wrote:

Dear Crystallographers,

I am trying to solve a structure in the space group P212121. Based on Matthews coefficient, there are 4 molecules in the asymmetric unit.

Based on my limited reading about using of Phaser, I understand that a single chain should be used as search model even though many copies are present in asymmetric unit. Am I correct?

So when I use a single chain as search model and ask Phaser to search for 4 molecules, Phaser identifies a single solution with a warning "The top solution from a TF rescoring did not pack" and a warning "Search request requires more scattering than defined in composition. Composition increased to accommodate search components". But the final values reported "PAK=2 LLG=1065 TFZ==22.6" indicate that phaser has solved the problem.

Can anyone please explain the meaning of the warning.

When I inspect the arrangement of the chains (attachment), I observe minimal contact between the chains and a large cavity in the center. Can a crystal form this way?

I have also tried using the dimer as search model and asking phaser to search for 2 molecules. Even in this case, Phaser finds a single solution but the warning and the advisory still appear as before. The numbers reported reduce a bit to "PAK=1 LLG=722 TFZ==29.2".

Please help me in understanding these results.

Thanks,

Satvik

 

 

 

<Monomer_searchmodel.pdf><Dimer_searchmodel.pdf>