[phenixbb] phaser spacegroup within phenix

[Randy J. Read]

Hi John,

Around the time of Phenix version 1.6.1, I think the default in the GUI was 
accidentally set to try all plausible space groups. As Nat said, in the 
current 1.6.2 release the default is to try both hands (for the few 
enantiomorphic space groups where you don't know which you have from the 
symmetry and systematic absences).

We've seen a few other cases of the problem you ran into. I'm guessing that 
there's pseudosymmetry, where the true space group is P21 but two molecules 
are related by a 2-fold parallel to the 2(1)? In that case, there would be 
some P2 pseudosymmetry. In such cases, Phaser can make a decision about 
space group prematurely. One of our plans is to change things so that the 
decision about space group doesn't have to be made with the first molecule 
but can be deferred until the difference in scores is unambiguous.

Regards,

Randy

On Jun 15 2010, John M. Berrisford wrote:

>Hello
>
>I've just spotted a discrepancy between phenix run from ccp4i and from 
>the phenix GUI regarding spacegroup selection run phaser runs.
>Within the CCP4 GUI you can specify the space group for a phaser 
>molecular replacement run to be the space group of the mtz or a 
>sub-space group (i.e P222 or P212121 etc...).
>For some reason this isn't as obvious within the phenix GUI.
>
>I have an mtz processed in P21 and when I load it into the phenix phaser 
>GUI it tells me its P21 (P 1 21 1) and gives me no other options. 
>However, when I run the molecular replacement job it decides that the 
>space group could be either P2 or P21 and runs both - which in my case 
>runs in P2 and gives a partial solution, not the full solution I would 
>expect in P21. A bit annoying as I know my space group is actually P21.
>After a few minutes of head scratching, the only way I found to fix it 
>to run in P21 is to enter the "other settings" menu and select "none" 
>within "alternative space groups". I guess this is what your suppose to 
>do, but its not obvious.
>
>Can the phenix GUI be changed so that its as obvious as the ccp4 GUI as 
>to what is actually going to run when you press "run".
>
>Thanks
>
>John
>
>I'm running phenix 1.6.1 on OS X.
>
>