or could it be in part a statistical effect, if the scala-map is
flatter, the differences near the ligand stand out more (i.e. are at
higher level relative to sigma), while they are really the same height
in absolute (electron-density) terms?
Mark van Raaij
http://web.usc.es/~vanraaij/
On 16 Sep 2008, at 18:56, Kay Diederichs wrote:
Hi Sabine,
difficult to say anything without knowing more about your procedures
and data. I would not exactly consider the difference between
20.9/24.8 and 21.6/26.1 as tiny.
One question - which procedure did you use to go from XDS_ASCII.HKL
to the mtz-file for SCALA? There are two different ways described at http://strucbio.biologie.uni-konstanz.de/xdswiki/index.php/Scaling_with_SCALA
but there may be other ways ...
What are the high-resolution R-factors of the model against the data
for (1) and (2)?
Maybe you could trace back the difference to a few high-intensity
reflections that were rejected as misfits by XSCALE, or as Wilson
outliers, using REMOVE.HKL, by you?
Are the positive and negative blobs at chemically sensible places,
i.e could the positive ones be waters and the negative ones mis-
placed model?
best,
Kay
Sabine Schneider schrieb:
Hello everyone,
I am puzzled about differences I see when I refine the very same
structure against data processed with xscale or scala.
I got data to 2.55A from a protein-ligand complex. The data were
processed with xds/xscale (1) or xds/scala (2)
Free R was imported from a previously solved structure with a
different ligand, first to (1) and from there to (2).
After MR and refinement (Refmac/Phenix), I get some difference
density peaks (a bit pos and neg), near the ligand when I refine
against (1) and it converges more or less around R/Rfree of 20.9
and 24.8. I do not get these pos and negative density blobs when I
refine the same coordinates against (2), with a tiny bit higher R/
Rfree of 21.6/26.1?
So the data processed with scala, where I get slightly better
refinement statistics, I end up with some pos/neg density peaks.
They don't show up when I apply a high resolution cut-off of 2.65A
during refinement.
Symmetry is orthorhombic and the statistics are OK (Rsym 0.09 and
0.38 in highest shell, redundancy ~5, I/sigI =1.9, mean I/sigI =
5.2 ) and in both files appear to be more or less the same number
of reflections.
I also did simulated annealing, omitting the region around the
ligand in a 15A radius and the density is nicely coming up. But it
doesn't make a difference: after refinement I always end up with
the result as discribed above? And there isn't really a difference
between the two structures in the end.
Has anyone an idea whats going on?
Sabine
