negative difference density around sulphur and oxygen atoms

From: Chris Meier
Date: 4 April 2012 16:16


Dear all,
I am refining the X-ray structure of a protein:
Data to ~2A were collected at a latest-generation synchrotron.
The 2fo-Fc maps are crisp, the model of the protein is complete and I am reasonably happy with the stats (R below 20%, Rfree below 25% in Refmac 5.5).
However, I am seeing a lot of negative difference density,
especially around sulphur atoms (negative density around -9 sigma)
and oxygen atoms (e.g. side-chain oxygens of Glu, Asp, etc. residues with negative density around -6 sigma).
Has anyone observed this before?
I have found CCP4bb postings discussing radiation damange of suplphur atoms
(e.g. http://www.dl.ac.uk/list-archive-public/ccp4bb/2004-07/msg00532.html ).
Can this also happen with oxygen atoms?
What would be an appropriate way to deal with this issue during refinement?
Suggestions greatly appreciated.
Thanks,
Chris


----------
From: Ian Tickle


Hi Chris

I would say there's something very wrong if you're seeing -6 sigma
difference peaks at O atoms.  I don't see how this can be explained by
radiation damage.  I for one have never seen that before in a
structure where there weren't other obvious issues (or maybe I just
haven't looked hard enough).

I would try refining it with a different program, e.g. Buster, or even
a different version of Refmac (I use 5.6.x routinely, but I see
there's a 5.7.x now - Garib will no doubt have an opinion on which is
the best one to use).  At least that will eliminate the software as
the origin of the problem: if it doesn't go away then we'll have to
think again.

Cheers

-- Ian

----------
From: Ian Tickle


PS you say the model is complete, but just as important how complete
is (are?) the data.

-- Ian

On 4 April 2012 16:16, Chris Meier  wrote:

----------
From: Roger Rowlett


Radiation damage induced loss of definition of disulfide bridges, side chain carboxylates, and certain histidine residues has been observed in synchrotron-irradiated protein crystals. For example, see Weik et al., PNAS 2000, 97, 623. I have also seen a recent paper where radiation damage of a bound protein ligand was apparently observed in a synchrotron beam.

I look forward to hearing from others how best to handle this in refinement.

Cheers,

_______________________________________
Roger S. Rowlett

----------
From: Jacob Keller

I look forward to hearing from others how best to handle this in refinement.

Dose-dependent occupancies (tau of an exponential decay function?) refined against unmerged data

JPK

*******************************************
Jacob Pearson Keller
*******************************************

----------
From: Scott Classen


Hello Chris,

Are you refining individual atomic B factors or grouped? Perhaps the B factors of the terminal atoms of the side chain are being restrained to too low of a B factor resulting in excessive negative density?

Scott

----------
From: Ian Tickle


The PNAS paper you refer to talks about a "loss of definition" of
exposed carboxyl O atoms, i.e. an increase in B factor, but presumably
if this is modelled properly then it shouldn't leave a big hole in the
difference map.  After all, the paper is not claiming that C-O bonds
are broken, only that there is "increased mobility" (or just as
likely, induced static disorder).  I'm wondering if this is related to
too-tight B-factor restraints.  I never use the default settings and
always use more relaxed ones: in particular I set the weights of B
factor restraints across angles to zero, IMO the across-bond
restraints are more than sufficient.  There has been a historical
obsession with getting B factors as low as possible (too-tight
restraints will certainly achieve this if that is your goal!), but
isn't the true goal of refinement to obtain the model which best
explains the data?

Cheers

-- Ian



----------
From: VAN RAAIJ , MARK JOHAN


apart from radation damage it could be a combination of:

- too tight restraints on the B-factors

- 9 sigma not being that much on a the e/A3 scale, i.e. your difference map is very flat (which is good) and the few peaks that remain stand out a lot, even if their absolute height is low...

Quoting Chris Meier:

Message

Mark J van Raaij

----------
From: Garib N Murshudov

Dear Chris

Could you please try later version of refmac then if the problem persists please let me know. Before making any suggestions it would be good to make sure that the problem is not related with particular software version (as Ian suggested)

regards

Garib

----------
From: Yuri Pompeu

could it be that the scattering table would be slightly different for the sulfur atoms at the collected wavelength?
Are they Cys or Met residues? if Cys is there a possibility of oxidation to the disulfides?

----------
From: Katherine Sippel <

On Wed, Apr 4, 2012 at 10:31 AM, Roger Rowlett  wrote:

 I have also seen a recent paper where radiation damage of a bound protein ligand was apparently observed in a synchrotron beam.

That was a manuscript were I would have happily given the coordinates and structure factors to the reviewers with my blessing. Learned a valuable lesson about adopting orphaned data sets though.

Cheers,
Katherine
 

----------
From: Sanishvili, Ruslan <

Hi Chris,

As has been suggested already, and seems quite plausible to me, it sounds like tell-tale signs of radiation damage.

To have little more substance behind this suspicion, some more experimental details could help:

What was the dose accumulated during data collection?

If the dose cannot be calculated, what was the beam intensity, frame exposure time, number of frames, total rotation of the crystal, crystal size, beam size? If the beam intensity is not known, the beamline and the attenuation factor used might be helpful.

 

What is the space group and how much data were collected? IF you have data with high multiplicity, you may be able to get rid of the latter parts of it maintaining completeness. This would reduce the effects of radiation damage if you are really dealing with it. Alternatively (but again with highly redundant data), you could try "zero dose extrapolation". Look up Kay Diederichs' and Dominika Borek's works on this.

 

Regards,

Nukri

Ruslan Sanishvili (Nukri), Ph.D.

----------
From: Eleanor Dodson


This could well be due to radiation damage - S are often affected, also Glu and Asp side chains. It is hard to know what to do since the effects are time related. If you have high redundancy maybe you could not use he later batches? Otherwise maybe just relax the B factor restraints and let them show the loss of atoms.. The trouble with that is that you have to relax all side-chain B restraints which may not be so appropriate for ILE say...
 Eleanor

On Apr 4 2012, Chris Meier wrote:

MessageDear all, I am refining the X-ray structure of a protein:Data to ~2A were collected at a latest-generation synchrotron.The 2fo-Fc maps are crisp, the model of the protein is complete and I am reasonably happy with the stats (R below 20%, Rfree below 25% in Refmac 5.5). However, I am seeing a lot of negative difference density, especially around sulphur atoms (negative density around -9 sigma) and oxygen atoms (e.g. side-chain oxygens of Glu, Asp, etc. residues with negative density around -6 sigma). Has anyone observed this before? I have found CCP4bb postings discussing radiation damange of suplphur atoms(e.g. http://www.dl.ac.uk/list-archive-public/ccp4bb/2004-07/msg00532.html ).Can this also happen with oxygen atoms? What would be an appropriate way to deal with this issue during refinement? Suggestions greatly appreciated. Thanks,Chris

--
Professor Eleanor Dodson