[phenixbb] Searchable phenixbb

Pavel Afonine PAfonine at lbl.gov
Fri Apr 11 14:11:39 PDT 2008

Hi Mark,

> I have an MR solution from a related MAD structure I solved and Arp/Warp 
> rebuilt 99% of this structure without any account for twinning.  

Just a remark... You could also do it in PHENIX using AutoMr, Autobuild 

> Upon refinement with twinning I get what I think are overly good (refined) 
> R/Rf values (particularily with waters = refine 3).  And refinement 
> without twining gives R/Rf on the bad side for my resolution (2A) but not 
> so terrible.  And these improve to "reasonable" R/Rf with the inclusion of 
> TLS and waters (= refine 5).  I have looked at the structures and maps for 
> both refine3-twin and refine5-notwin and see little differences but 
> nothing major.  So my dilemma is which structure to continue refiniing the 
> twinned or not twinned+TLS, how do I know which is correct?

Looking at the numbers, I see no surprises: the most reasonable strategy 
produces the most reasonable results: refine5 (+TLS+H2O) with 0.1428 / 
0.1770 for R-factors.

- Yes, definitely accounting for twining is important since it seems to 
be present and it is supported by the R-factors.
- At 2A resolution we normally see many waters, so you need to ask 
phenix.refine to keep them updated all the time using 
"ordered_solvent=true" option. The set of water picking parameters gives 
you a full control over the process, so you can adjust the amount and 
quality of picked waters.
- The total atomic B-factor in phenix.refine is defined as :
Btot = Bcryst + Btls + Blocal, where Bcryst comes from overall 
anisotropic scaling, Btls models the global motion of the domains or 
molecule as a whole, and Blocal models local atomic vibrations.
If you use TLS you utilize a better model for the atomic ADP (providing 
a correct choice of TLS groups -- it is very important!), and without 
using TLS you force Blocal to model it which is not great (especially 
since the similarity restraints are used for ADP). So I would use TLS, 
but make sure to selects reasonable TLS groups.

> My other questions alternate conformations, is it possible to 
> refine the occupancy?

You can do it with the latest version available from 
however it has a bug that will be fixed in the next release (in a few 

Please let me know if you have any other questions!

More information about the phenixbb mailing list