Refining occupancies with autoBUSTER

Here, we're using the recently-released structure 2qp6, in which there are two mercury atoms, one modelled as a single Hg++ ion and the other as mercuribenzoic acid.

Seeing what's going on

An initial refinement of this structure with

refine -p 2qp6/2qp6.pdb -m 2qp6/2qp6.mtz -nsmall 10 -nbig 2 -d initial

gives nine difference peaks of worse than ten sigma, concentrated in two areas: residue MBO465 2qp6-MBO465-1.png

and residue HG464

2qp6-HG464-1.png

It's pretty clear what's going on at the bottom: the mercury is on an MBO ligand, and the occupancy needs to be adjusted. For the top site, turning off the (confusing) difference density and contouring at 7-sigma suggests that the mercury might be well-modelled as two conformations, with the SG of the cysteine moving left into the difference density there, and one of the mercury conformations centred in the green blob to the left of the present mercury conformation.

There is some extra difference density around the MBO465 site, so initially we get rid of the MBO and model only the two mercuries there (as conformations A and B of a residue HG C1), while deleting the waters in the MBO-density near HG464 and replacing HG464 itself with a new residue MBO D1 using coot. This gets us 2qp6.1.pdb

Performing the first occupancy refinement

Now we need to write a Gelly file to tell autoBUSTER what to refine: this is 2qp6.1.gelly

NOTE BUSTER_RESET_CONSTANT_COMBINE
NOTE BUSTER_SET Mercuries = {C|1:*.A C|1:*.B D|1:*}
NOTE BUSTER_SET EverythingElse = \ Mercuries
NOTE BUSTER_COMBINE OCC {D|1:*}
NOTE BUSTER_CONSTANT OCC EverythingElse

An explanation of the five lines:

  1. Reset the refinement state; by default occupancies are all fixed to their initial values, NOTE BUSTER_RESET_CONSTANT_COMBINE turns this off. It also turns off any grouping of positions or B-factors, but at this resolution level that's desirable.
  2. Define a set containing the two alternate-conformation Hg atoms and the MBO residue
  3. Define a set containing everything else
  4. Arrange that only one occupancy is refined for the atoms in the MBO residue
  5. Hold the occupancies of everything else constant

To run a refinement using this Gelly file, do

refine -p 2qp6.1.pdb \
       -Gelly 2qp6.1.gelly \
       -m 2qp6/2qp6.mtz \
       -d run1 | tee run1.lis

The resulting electron density around the two mercury sites looks like

run1-465.png

run1-464.png

So, we have a very nice fit for the 464 site, but still a lot of negative difference density on the mercury. Anomalous scattering and occupancy issues can easily be confounded, and mercury is a strong anomalous scatterer at almost all wavelengths of crystallographic interest, so it's probably worth following the process in AutoBusterExampleFormfactor and adding

FormfactorCorrection="Hg:-6.7"

(this being the figure crossec gives for mercury's f' at 1.2A) to the refinement command line.

Performing a second occupancy refinement

Let's restore the benzoic acid at the site near 465, and refine the occupancy of the mercury in D1 separately from the benzoic-acid part of the residue.

So, the atom which is now C1 HG:HG.B should become E1 MBO:HG.B, and the atom which is now C1 HG:HG.A will stay C1 HG:HG.A; the other atoms in E1 MBO all get alternate-conformation code B, and we write a slightly more involved .gelly file

NOTE BUSTER_RESET_CONSTANT_COMBINE
NOTE BUSTER_SET MercuryContainers = {C|1:*.A D|1:* E|1:*.B}
NOTE BUSTER_SET EverythingElse = \ MercuryContainers
NOTE BUSTER_SET MBOD1 = {D|1:*}
NOTE BUSTER_SET MBOD1_Hg = {D|1:HG}
NOTE BUSTER_SET MBOD1_other = MBOD1 \ MBOD1_Hg
NOTE BUSTER_COMBINE OCC MBOD1_other
NOTE BUSTER_COMBINE OCC {E|1:*}
NOTE BUSTER_CONSTANT OCC EverythingElse

and run

refine FormfactorCorrection="Hg:-6.7" -p 2qp6.2.pdb -Gelly 2qp6.2.gelly -m 2qp6/2qp6.mtz -d run2 | tee run2.lis

It is important to note that combining occupancies for a residue whose atoms start off with different occupancies means that all the occupancies are adjusted by the same amount - which means that it is vital not to insert the organic part of a metal-containing ligand with the occupancy at 1.0 and the attached metal atom's occupancy already refined, since the metal atom's occupancy cannot then be adjusted upwards. It's best to set the occupancies for the whole organometallic complex to 1.0 and let the optimiser sort things out.

After this run, the worst density features are no longer on the mercury atoms.

run2-464.png

run2-465.png

The model isn't perfect - there are clear features on the bottom MBO in the 2-positions either side of the mercury, but there are no huge density peaks corresponding to the mercury if we thought of those as altconfs of the whole MBO - but it's significantly improved.

The mercury-carbon bond is not that strong (2.07A bond length) and so is very prone to radiolysis; it might be nice to model one conformation as MBO, and another conformation as a mercury bonded to the SG of the cysteine and coordinated to the carboxyl group of a BEZ located at a sixty-degree rotation to the current BEZ.

Page by Tom Womack original version 9 June 2008. Address problems, corrections and clarifications buster-develop@globalphasing.com