Refmac and metal on a two-fold?

From: Dima Klenchin
Date: 28 December 2011 19:29

Hello,

I have a metal ion sitting on a two-fold. I assigned it an occupancy of 0.5 but Refmac keeps refining it away from that position so that in the end there is a symmetry-related ion 1.6A away. I thought that this problem has been rectified long ago? Certainly I had several occasions where Refmac would automatically recognize special position and automatically assigned proper occupancy (1/n). Not sure why this particular case is different...

This behavior is the same in:

Refmac 5.5.0088 (Windows)
Refmac 5.5.0102 (Windows)
Refmac 5.6.0117 (Windows)
Refmac 5.5.0109 (Linux)

I also tried leaving it at occupancy 1.0 and also tried renaming MN2+ to water at occupancies of 0.5 and 1.0. All to no avail - the atom is always shifted away from the density.

Anything I can do to force Refmac to keep it nice and tidy?

Thanks and Happy New Year!

- Dima

----------
From: Ed Pozharski


I've seen this happening to water molecules as well (in a somewhat
unpredictable fashion).  In the latest refmac versions, you can try
harmonic restraints, although these will only slow down the atom drift,
as the target position is updated every cycle.

Perhaps you can use distance restraints against a dummy atom to fix the
metal ion in place.

Another remedy might be to manually place the metal ion in position
followed by zero cycles of refinement to get the final R-values.  There
is some evidence that excluding isolated atoms from refinement may
introduce some systematic error on the rest of the structure, but it is
likely very small.

Cheers,

Ed.

----------
From: Dima Klenchin


Thanks, Ed!

Just in case anyone else has the same issue:

With Garib's help, I have forced the atom into its position by using external distance restrains of zero length against the same symmetry-related atom. The cause is unclear because the same program handles special positions in another structure just fine. Here is the exact command I used:

EXTERNAL DISTANCE first chain M residue 3 atom MN -
  second chain M residue 3 atom MN value 0.0 sigma 0.00 symm Y

Occupancy set manually to 0.5.

- Dima

----------
From: Cale Dakwar

Good to know, thanks.  I was actually wondering about this recently.

----------
From: Ed Pozharski


Thanks - good to know.  Perhaps a good practice is to take care of
special positions manually every time just in case.

>From what I understand, the way refinement software generally treats
special positions (other than automatically resetting occupancy to 1/z)
is to exclude vdw repulsion to symmetry-related atom which in this case
would always push the atom out of position.  Exclusion is usually
applied to the atom when it is within the cutoff distance (0.1A?) from
the corresponding symmetry element.  If for whatever reason the atom
shifts, the vdw will push it out.  Why would such atoms shift varies - I
suspect in some cases it is due to hydrogen bonding restraints, in
others it could be due to noisy density.



--
Oh, suddenly throwing a giraffe into a volcano to make water is crazy?
                                               Julian, King of Lemurs

----------
From: <Herman.Schreuder


Dear Ed,

I thought the clumsy way of handling special positions you describe was
abandoned after the Xplor days, that even rounding errors could cause
special atoms to fly apart, but you may be right. The Mn ion might not
really be special, but might have two minima, one closer to one protein
molecule and one closer to the symmetry related protein molecule.
Supposedly, refmac switches off vd Waals repulsions if the combined
occupancies of the the atoms involved does not exceed 1.0. This should
take care of the problem but apparently does not. Forcing the ion to
have to same position as its symmetry mate may not always be the best
representation of the reality in a crystal.

The main reason I switched to Buster a few years ago is that in refmac
residual repulsions remain, even if the summed occupancies of atoms in
alternate positions do not exceed 1.0.

Best,
Herman

----------
From: George M. Sheldrick


I am surprised that this is such a mess. Small molecule
crystallographers frequently encounter special positions and have been
able to handle them correctly for many years. A well known small
molecule refinement program that I don't need to advertise here has been
able to generate the special position constraints for the atom
coordinates and anisotropic displacement parameters fully automatically
for all special positions in all space groups for the last 20 years, no
special user action is required.

George

--
Prof. George M. Sheldrick FRS

----------
From: <Herman.Schreuder


Dear George,

I think for small molecules the situation is much easier as for
proteins. With small molecules, the resolution of the data is in general
much higher and a water or metal ion will force the small molecules in
such a position that the special position is exact. With proteins, the
water or metal ion will be in one of many crystal contacts and may not
be able to force the much larger proteins in such a position that the
special position is exact and what we think is a special position is in
fact a mixture of two or more alternative positions. I strongly believe
that in these cases one should not force the water or metal ion to be
exactly on a symmetry axis, but just switch off the vd Waals repulsions
and let the data decide which water or metal ion positions best
corresponds with the diffraction data. This switching off of vd Waals
repulsions does not seem to always work with refmac.

Best regards,

Herman