[phenixbb] refinement with translational ncs

[Jorge Iulek]

Dear all,


     I have a dataset at 2.98 A resolution, images processed to CC_1/2 
at circa 50% in the last resolution shell, tetramer in the asymmetric 
unit, local (rotational) symmetry 222. Xtriage indicates a peak in the 
Patterson function at 20% height of origin peak, at fractional 
coordinates of 0.0000 0.0000 0.3177. Space group is P21.

      I am aware that this brings problems to refinement , as even 
stated at xtriage output  "Translational pseudo-symmetry is very likely 
present in these data.  Be aware that this will change the intensity 
statistics and may impact subsequent  analyses, and in practice may lead 
to higher R-factors in refinement.", and so I can't lower my R and 
R-free below ~31 and 36 % , respectively.

     At phenix.refine, setting "tncs_correction = True" brings little 
benefit (what does this actually do? points to references are welcome). 
I wonder nevertheless if some strategy might help in on lowering these 
R's and, of course, allow me to build a better model (in spite of the 
resolution). I tried some combinations of other types of ncs and tls 
groups, with little success. Particularly, it seems that the refinement 
of ADPs are a kind of unstable (what I would understand come from the - 
partial - modular nature of the intensities - due to tncs - and these - 
ADP - would be reflected by the intensity fall-off with resolution). 
What about refining ADPs only to a zone which is not affect by data 
modulation? How to do that? Any other ideas?

     Thanks,


Jorge