[phenixbb] twin/pseudo-symmetry question

[Nathaniel Echols]

On Fri, Nov 11, 2011 at 10:03 AM, Green, Todd <green at cbse.uab.edu> wrote:
> For the R32 set, I generate the r-free in phenix directly from the scalepack file for data processed in R32. For C2, it was generated from the C2 processed scalepack file but in ccp4 using the scalepack2mtz script then imported to phenix(because i used molrep for the molecular replacement). But, I kept this same data file with the same r-free for all of the refinement runs before and during twin refinement. Seem reasonable?

I'm not totally certain about this, but I don't think scalepack2mtz
will take the higher lattice symmetry into account.  When you generate
the flags in Phenix, by default reflections which are related by
possible twin laws will have the same flag.  If you don't do this,
however, when you use twinned refinement you will effectively be
refining against (most of) your test set, which leads to an
artificially low R-free.  (I'm probably butchering the terminology
here, sorry.)

What I would do instead is take the MTZ file from scalepack2mtz and
get rid of the R-free flags, and let Phenix generate new ones - then
start refining the MR solution with the new file.

-Nat