*** For details on how to be removed from this list visit the ***
*** CCP4 home page http://www.ccp4.ac.uk ***
> Give refmac a proper set of scattering factors for the wavelength
> your data was collected at. This is controlled by the input file
> name ATOMSF. The default file is in ${CLIB}/data/atomsf.lib
> But the contribution of a metal atom at the inflection point of
> its absorption edge really is smaller, so even with the correct
> scattering factors in place you should still expect a lower peak
> in the maps made from this X-ray energy.
I was asked off-list if the failure to replace scattering
factors in atomsf.lib could also explain negative electron
density at Se sites in maps made for SeMet proteins.
The short answer is "yes, but..."
The longer answer continues...
We have many SeMet structures in the lab for which there are
negative peaks at the Se sites after conventional refinement.
I believe that there are several contributing factors.
(1) As per the query on Mn++ scattering, the default f' scattering
factor in ${CLIB}/data/atomsf.lib is correct at 1.54A but not at or
near the Se absorption edge. To be correct, you should change the
value of f' for Se to something on the order of -9e at the inflection
point. You should also have an experimentally determined value for
both f' and f" from an edge-scan performed on your crystal at the
time of data collection.
(2) Incomplete incorporation of SeMet (some of the Se atoms are
really S atoms). This can be quantified by mass-spec if you really
care.
(3) Loss of the Se atoms due to radiation damage. Split your data
into two halves based on time of collection. Is the density
more negative in maps calculated from only the second half?
(4) Disorder of methionine sidechains. In my experience, very high
resolution structures show that most Met sidechains have multiple
conformations. I'm sure this is true at lower resolutions as well,
but typically it is not so blatantly obvious.
(5) Too-tight B-factor restraints. Methionine is a relatively
long amino acid. You could try loosening the sidechain B-factor
restraints from their default values.
Bottom line: except in extreme cases, negative density at the
Se sites should not worry you too much. Arguments (2), (3),
and (4) above could be used to justify assigning partial occupancy
to the Se atoms (in which case you should for consistency probably
do the same for CE).
Probably the most conservative treatment is simply to
acknowledge the problem by stating "the largest features remaining
in the residual Fo-Fc density map after refinement were negative
peaks at the Se sites". And if that isn't, in fact, true -
well then you really should be looking at those other difference
peaks before wasting time on fiddling with the Se descriptions :-)
--
Ethan A Merritt
Biomolecular Structure Center
University of Washington, Seattle WA
