[phenixbb] SA omit map

[Nathaniel Echols]

On Wed, May 30, 2012 at 11:56 AM, Huang, Raven H <huang at illinois.edu> wrote:
> Was the issue of SA omit map debated last July (involving Tiaard Pining,
> Pavel, and Folmer) resolved in the newer version of Phenix? Specifically,
> We'd like to generate the same map for a ligand. It is not for the
> refinement, but for publication (a reviewer asked us to cover the ligand
> with the map in a figure).

After reading over the original thread
(http://www.phenix-online.org/pipermail/phenixbb/2011-July/017342.html)
I don't think there was anything to resolve - the issue was a
misunderstanding about what the output from the omit map calculation
actually includes.  The MTZ file produced by Phenix will cover the
entire unit cell, and if you convert this to a CCP4 map, the coverage
depends on the program settings.  (If you open the map in PyMOL from
the Phenix GUI, it will cover the entire asymmetric unit plus
padding.)  If the map type is a 2mFo-DFc map, the interpretable
density will naturally include the protein.  If it's an mFo-DFc map,
the ligand should stand out.  In either case, if you want to
explicitly exclude any non-ligand atoms from the density that appears
in your figure, you have several options:

1. In phenix.refine, you can tell it to output pre-calculated CCP4
maps covering only an atom selection.  However, I think the result
will be a box with boundaries defined by the extents of the atom
selection, which may still include protein atoms.
2. phenix.cut_out_density (also in the GUI) does exactly what it
sounds like; I think this will mask around the atom selection.
However, it will also shift the extracted density to a P1 box (along
with the model), which may not be what you want.
3. PyMOL probably gives you the greatest amount of control, via the
isomesh command.

Personally, I would take the mFo-DFc map from the last refinement
before you added the ligand, and show that.  Otherwise, just run
phenix.refine with SA using the ligand-free model, and load the
mFo-DFc map.  (Unless the reviewer wants the 2mFo-DFc map, but with
the mFo-DFc map you can usually show just what you're interested in
without too many tricks.)  I wouldn't spend a huge amount of time
trying to make the density only cover the ligand with no visual
artifacts - this can be very misleading and doesn't really make the
figure any clearer.  (In fact, in at least one case I wished that the
authors had shown the density for the protein too, because I had no
way to judge the quality of their ligand density by itself.)

Maybe this needs to be added to the FAQ list?

-Nat