Dear all,

a bit of a background follow-up to this:

On Tue, Jan 14, 2025 at 09:37:44AM -0500, Yong Tang wrote:
> I do understand PDBe offers electron density maps (8hbk.ccp4 and
> 8hbk_diff.ccp4 separately)- is that our go-to nowadays when we would like
> to inspect the maps of any particular PDB entry
             ^^^

The tricky part is if you want to fetch "the" map or just "a" map ;-)

As discussed by others in that thread, for a given PDB entry we might have

 (1) the maps as provided by the depositor (most likely as computed by
     the refinement program used/reported)

     * electron density maps a la 2mFo-DFc (possibly in variants of
       DFc completion)

     * difference maps (mFo-DFc)

 (2) maps provided by wwPDB sites (map coefficients as MTZ from PDBj,
     CCP4 maps from PDBe and map coefficients as CIF from RCSB)

     * when map coefficients are provided they seem to use isotropic
       DFc completion (i.e. all unobserved refelctions to the
       diffraction limit of the observations use DFc instead of
       2mFo-DFc in the electron density map computation).

     * the CCP4 map files contain no information how they were
       computed (DFc completion or not)

 (3) maps recalculated by user from deposited reflection data (with a
     program of their choice and a specific handling of missing
     observations) - where the deposited reflection data might or
     might not be completely identical to the data that went into the
     final refinement step (and produced the maps from the depositor).

All of these might provide maps that will look different (grid,
handling of missing observations, bulk solvent methods, scaling,
outlier handling etc) ... so tricky to define one as "the" map.

> https://www.ebi.ac.uk/pdbe/entry/pdb/8hbk

Taking this as an example and looking at the electron density maps
available:

 * using the PDBe-provided map files to define the grid (otherwise a
   comparison is tricky)

 * converting CIF files into MTZ using "gemmi cif2mtz"

 * computing maps (on same grid as PDBe maps) using one of FFT, CFFT
   or gemmi (they all come up with the same maps - good)

This shows for the 2mFo-DFc map (FWT/PHWT columns in CIF/MTZ) the
following stats (using "gemmi map"):

  Source     Min      Max      Rms
 ------------------------------------
   PDBe   -0.58902  1.87719  0.19755
   PDBj   -0.56885  1.90967  0.19660
   RCSB   -0.56885  1.90967  0.19660

Since one doesn't know if the PDBe maps used DFc completion for the
unobserved reflections, we can also compute maps for the observations
only:

  Source     Min      Max      Rms
 ------------------------------------
   PDBe   -0.58902  1.87719  0.19755
   PDBj-2 -0.58976  1.89473  0.19603
   RCSB-2 -0.58974  1.89469  0.19602

That shows that map files from PDBe are different from maps computed
using the map coefficient files provided by PDBj and RCSB (and those
two are basically identical to each other - probably based on the same
underlying validation software). We can also comparte these to the
maps computed from the deposited map coefficients (i.e. the FWT/PHWT
columns in the deposited reflection data mmCIF file) ... and get a
large list of slightly different maps:

  Source             Min      Max      Rms    Remark
 ----------------------------------------------------------------
   PDBe          -0.58902  1.87719  0.19755   ? completion ?
   PDBj          -0.56885  1.90967  0.19660   DFc completion
   PDBj-2        -0.58976  1.89473  0.19603   obs only
   RCSB          -0.56885  1.90967  0.19660   DFc completion
   RCSB-2        -0.58974  1.89469  0.19602   obs only
   Deposition    -0.59226  1.85841  0.21428   DFc completion
   Deposition-2  -0.60827  1.84176  0.21368   obs only

That table would probably get even longer if one would use a favourite
program to compute those afresh: I'm sure PHENIX, REFMAC and BUSTER
(and different versions of those) would all create slightly different
maps.

Now what is "the" map to look at? One could argue that

 * "The" map for 8hbk when looking at that structure alone (and maybe
   the accompanying paper) is the one provided by the depositors,
   i.e. computed from the deposited "structure factor" file.

 * "The" map when looking at 8hbk and a whole bunch of other PDB
   structures could be the one using the map coefficients from PDBj or
   RCSB (since they are nearly identical): the CCP4 maps provided by
   PDBe might be more practical, but it is a bit concerning that they
   seem to differ from the offerings of the other wwPDB sites.

Does it all matter? Hopefully not: those differnces should be rather
small (but if they are not I would start getting worried) - as long as
one is aware of intrinsic differences between (1) deposited maps (from
wwPDB entries), (2) the look-and-feel if using a different grid, (3)
effects of DFc completion approaches and (4) different algorithms in
different programs computing map coefficients afresh.

Cheers

Clemens

########################################################################

To unsubscribe from the CCP4BB list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1

This message was issued to members of www.jiscmail.ac.uk/CCP4BB, a mailing list 
hosted by www.jiscmail.ac.uk, terms & conditions are available at 
https://www.jiscmail.ac.uk/policyandsecurity/

Reply via email to