*** For details on how to be removed from this list visit the ***
*** CCP4 home page http://www.ccp4.ac.uk ***
As is often the case with these types of discussions I feel we are all
more or less thinking the same. At 3.5 A the experimental data provides
only weak support to assign waters and you'll have to work hard to
convince yourself, and harder, to convince the reviewers/readers that
there really is a water. This is worth the effort when you are focusing
on some important waters that are likely relevant for function but
assigning waters based on a peak-list from a difference map and
accepting all that don't clash with the protein and make at least one
hydrogen bond does not sound like good practise. In other experimental
fields you typically use the 5% rule where something is significant only
if the probability that you are wrong is less than 0.05. Placing waters
is a much less quantitative activity but if you apply that stringency
then you will need significant support from other sources of information.
That is why Jack's use of conserved water sites in families of related
structures is a good development. In past studies waters were found to
be poorly conserved even across structures of the same protein in
different crystal forms, so if some water sites are conserved between
several related structures then those are the waters on which you want
to spend your time to see if they are present in your low resolution
structure.
Bart
Mischa Machius wrote:
*** For details on how to be removed from this list visit the ***
*** CCP4 home page http://www.ccp4.ac.uk ***
I would be very suspicious of "general rules" derived from an analysis
of low-resolution structures found in the PDB. If that analysis says
"people don't place waters into their structures at 3.5Å", then that
does not necessarily mean that it is never justified to place waters at
3.5Å. As Martyn rightfully pointed out, that analysis may say more
about "people", i.e. model building and refinement practices, than
about objective physical reality (whatever that is).
I think the issue really is that somebody is observing a feature in the
electron density at 3.5Å and attempts to explain that feature. To treat
that situation in a rather formal way, the first step would be to
create a list of (plausible) hypotheses for an explanation. Let's
assume that a water molecule could satisfactorily explain that density
feature. When is it "allowed" to include the water molecule in the
model, and when not? Clearly, crystallography on its own, particularly
at a resolution of 3.5Å, is not really suitable to _prove_ that a
feature is a water molecule and not an ion or an FFT ripple or noise,
or something else. Additional information is needed to increase the
confidence that the feature is indeed a water molecule. For example, if
there are similar high-resolution structures that "show" a water
molecule, then I think it is perfectly legitimate to include that water
molecule in the model for the 3.5Å structure. After all, that is how
many features are interpreted at that resolution. Many times, people
brush features aside, saying that's just "noise". What is the objective
justification for that?
Model building and refinement make use of a lot of external information
about structural properties of amino acids, nucleic acids, etc., even
at very high resolution and even more so at low resolution. I think we
can just as well derive some "general" rules from _high_resolution
structures about waters and use those rules to interpret low-resolution
structures, just like Jack alluded to in an earlier post.
That does not mean one should now drown all low-resolution structures
with waters. It only means that things are never really simply black
and white. They often are green and blue and red...
--
==============================================================================
Bart Hazes (Assistant Professor)
Dept. of Medical Microbiology & Immunology
University of Alberta
1-15 Medical Sciences Building
Edmonton, Alberta
Canada, T6G 2H7
phone: 1-780-492-0042
fax: 1-780-492-7521
==============================================================================
