[phenixbb] Restraining/constraining partial occupancy waters

David Briggs drdavidcbriggs at gmail.com
Tue Oct 27 06:56:32 PDT 2015

Hi PhenixBB.

I have a very minor problem with partial occupancy waters moving out of
density during refinement

I have 2 structures, one apo structure (~1.3Å), one ligand bound (~1.5Å).

The ligand binding site contains a Calcium ion. In the apo form there are 2
waters co-ordinating this Calcium. In the ligand bound form, these waters
are replaced by oxygen atoms within the ligand.

However, the ligand is only at 0.5 Occupancy as part of the ligand lies on
a crystallographic two-fold axis. The ligand itself refines well with
occupancy fixed at 0.5, and the Bs are sensible.

In the 50% of the molecules which don't co-ordinate ligand, two waters must
co-ordinate the calcium ion, and in keeping with this, we see that the
electron density around these positions is greater than would be expected
for the 50% occupancy ligand. So I added 2 0.5 occupancy waters to the
model at these positions.

However, when I run phenix.refine (refining XYZ, real space, occupancy, and
individual Bs (isotropic for the ligand & waters - for the protein I use
anisotropic Bs) these two 'ghost waters' move out of the electron density
and their Bs increase massively.

*My questions are:*

Is there any way to restrain these waters to their apo positions? Is this
the correct approach?

I had initially thought that reference model restraints might be useful
here, but if I understand correctly, reference model restraints cannot be
applied to waters - Is this correct?

If I cannot restrain the waters to the apo positions, would I be justified
in fixing (constraining) them? This seems a little extreme and I am loath
to do it, but given the isomorphous apo structure, I do have evidence that
in the absence of ligand waters do take up these positions.

Can anyone help?



[image: David Briggs on about.me]

David Briggs
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://phenix-online.org/pipermail/phenixbb/attachments/20151027/bb8f5333/attachment.htm>

More information about the phenixbb mailing list