Lieven Buts wrote:
Hello all,
I have just been experimenting with the refinement options for twinned data
in phenix.refine (final 1.3 version), and I ran into a strange problem with
the density fit validation function in Coot 0.4 and 0.5. When I "auto-open"
the MTZ file with map coefficients produced by phenix.refine, I get a
normal-looking map, but every single residue in the density fit plot is
squarely in the red. When I open the 2Fo-Fc CNS format file also produced
by phenix.refine, I get a map that looks identical to the first one, and
when I validate the same model against the new map, I get the normal
density fit pattern, with most residues in the green, and the occasional
orange one. So it seems that the density fit calculation goes wrong for
MTZ files from twinned refinement. Does anyone know what might be
causing this? I can send the data files if required.
Randy Read told me about this issue the other day.
The density fit graphs work assuming that the map is on the absolute
scale - I suspect that your maps have a different scale (test by looking
at the absolute level given the same number of sigma). It would be
better if Coot used correlation vs a calculated map, but today it does not.
You can adjust the scale using Extensions -> Refine -> Set Density Fit
Graph Weight [1]
[1] I may move this soon - Settings would seem a better submenu.
Paul.
