[phenixbb] cut out density vs FFT

[Nathaniel Echols]

On Tue, Feb 19, 2013 at 11:06 AM, Yunji Wu <wuy at caltech.edu> wrote:
> My questions are (1) what should I use for what I am trying to do (just display 2Fo-Fc density around asymmetric unit for a figure) and (2) what is the difference between what these three programs (cut_out, FFT-CCP4, FFT-Phenix) are calculating? The inputs are exactly the same -- mtz (2FOFCWT and PH2FOFCWT) and pdb output from refinement.

To answer the first question, I think it's sufficient to do one of two things:

1) carve out the mesh in PyMOL (or preferred graphics program) so it
only covers the protein - you may have to experiment with the
'isomesh' command to find the perfect settings.  You may not get
perfectly rounded mesh where it extends beyond the protein, but at
this resolution I doubt it will detract very much from the figure.

2) don't bother hiding the density outside the ASU, just show where
the ASU is by displaying the cartoon model.  This is what I've ended
up doing in similar circumstances (for instance:
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3382516/, Figure 3).  It
is mostly a matter of personal preference; I am very paranoid about
phase bias and I've seen too many dishonest figures (e.g. 0.2 sigma
density around a ligand claiming to be 1 sigma density, but without
showing density for the nearby protein atoms).  What you lose in
aesthetic quality, you gain in realism.

There are several differences between the programs themselves.  I do
not know what the default gridding is in CCP4; most of the time when
we output a .ccp4 map it is using d_min/4.  Another subtle issue is
how big of a box around the protein to cut out, and how this affects
the distribution of values and thus the meaning of the different sigma
levels.  Depending on how much bulk solvent you throw out, the density
at "1 sigma" may change volume.  Finally, I don't know whether
cut_out_density does any kind of interpolation of values (like
phenix.superpose_maps) or whether it instead preserves the original
grid - if the former, this can conceivably lead to artifacts.

Note that phenix.cut_out_density was really intended for preparing
maps for use in MR - although I suspect many people end up using it
for display purposes.

All that aside, I'm a little concerned if the programs are producing
that different output.  Could you please send me the input files
off-list so I can check whether something isn't working as intended?

-Nat