XPLOR and REFMAC

[Yu Wai Chen]

Dear all,

I am quite confused in a few aspects concerning shuttling between XPLOR
and CCP4 REFMAC.

I am trying to refined a structure, with 1.8A data, solved by molecular
replacement (CCP4 AMORE) with a unambiguous solution.

First I tried X-PLOR: rigid body, SA, torsional dynamics, positional
refinement, with and without bulk-solvent correction, none of these
produce a drop in Rfree to below 44%.

I then tried restrained REFMAC, with anisotropic scaling, bulk solvent
correction, and individual B-refinement.  I can then get good R/Rfree =
24%/30%.

At this stage, I am still lacking about 13% of the protein (a loop) so I
would like to build SA-omit maps (on recommendatin of Alwyn Jones) to
try to find it.  So I take the model and the data (with missing
reflections filled in and free-set set aside by CCP4 uniqueify), convert
the data to X-PLOR format and try to build the maps in X-PLOR.

However, I observed very bad R/Rfree (50%/53%) when I ran the
check.inp.  I noticed that the B-factors of the previously
X-PLOR-refined structures (R/Rfree~40%/44%) are considerably lower than
the REFMAC-refined structure.  I understand that if X-PLOR is used
without bulk solvent correction, it is underestimating the B-factors and
so the higher B-factors in REFMAC structure is actually more realistic. 
So I tried to set up a bulk-solvent-corrected data set in X-PLOR but I
get no improvement in refinement.  I then experimented with refining
individual B in X-PLOR starting with the REFMAC-refined model.  I get
much lower R/Rfree = 40%/43% this time.  And the B-factors are "refined"
back to very low values (most below 15A2).  But still these R/Rfree is
way higher than what I get in REFMAC.

So I have a few questions:
1. I suspect it is the anisotropic scaling which is doing a great help
in lowering the R/Rfree in REFMAC.  Is there a way to do this in X-PLOR?
2. Shall I use the maps calculated with a SA procedure that results in
R/Rfree of 47%/50%; or shall I also add individual B-refinement to this
SA-omit map procedure and use maps that ends in lower R/Rfree?
3. What is the best way to shuttle between these 2 programs and see
similar R/Rfree results?

Thanks for your input in advance.

-- 
===================================================================
Yu Wai CHEN, Ph.D. ..................  email: ywc at mrc-lmb.cam.ac.uk
 Centre for Protein Engineering,              tel: 44-(1223) 402148
 MRC Centre, Cambridge  CB2 2QH, U.K.         fax: 44-(1223) 402140
 WWW homepage: http://www.mrc-cpe.cam.ac.uk/people/wai.html