Back to Taiana - it is almost impossible to give a definitive answer to
this.
a) Different software gives diff estimates of Sigma - try comparing
HKL2000 and MOSFLM/Scala - the Inensities will agree very closely but
prob. not the Sigma estimates. These are in all cases fudge factors, and
not terribly reliable!
b) Different crystal quality affects the apparent limits. eg if your
data is very anisotropic reporting I/Sigma in resolution shells isnt
very useful - you may have excellent observations in one zone and
nothing in the perendicular direction. But it is still important to use
the maximum resolution you can reach..
c) It depends what you are planning to use the data for. For exptl
phasing ( SAD/MAD or MIR) you need a decent signal for a good % of the
meaasurements - prob I/SigI > 3.
But for refinement it is always best to use all the observations you can
get. A local convention is not to integrate past the point where you can
no longer "see" some spots in an image, with all the magnification etc
available.. For a standard crystal that means I/Sigma is ~ 1.0 or a
bit lower, but that cutoff cannot be a definitive one.
Eleanor
b) Juergen Bosch wrote:
*** For details on how to be removed from this list visit the ***
*** CCP4 home page http://www.ccp4.ac.uk ***
Sorry Bart, that was not clearly phrased from my side.
What I tried to say is during refinement I use as much data as
possible, then I cut back the resolution and rerefine, then table1 is
generated and the extra 0.2 A which were cut off are additionally
deposited for e.g. developers.
I hope that is clearer now ?
I just think that all recorded data should go to the PDB - I might be
alone with my opinion but I think it actually might be useful to
others. And yes all structures are deposited with structure factors &
x wavelength if it was a MAD or SAD data set.
Sorry for causing confusion.
Jürgen
Bart Hazes wrote:
Juergen Bosch wrote:
I agree with Bart in the "reviewing part" but during refinement I
use all available data, depending on NCS, completeness and
redundancy it might also be lower than 1 sigma but not lower than
0.5. I would expect that XDS will have a better I/sigI versus e.g.
HKL2000 - but maybe somebody much wiser might comment on this.
Once the refinement is done Table 1 is adjusted to meet the
expectations of reviewers I/sigI and Rsym, but the stuff that goes
to the PDB will contain all data :-)
Maybe I don't understand you correctly but the role of Table 1 is to
describe the experimental data that was used to derive the model. You
can't use resolution X for refinement and then report statistics up
to resolution Y because it looks nicer. Reviewers should definately
have a problem with that procedure. If you think the model is better
by following your procedure, and I'm not saying it's not, then you
should be able to state it in the paper and refute any comments
reviewers may have.
Bart
Jürgen
Bart Hazes wrote:
*** For details on how to be removed from this list visit the ***
*** CCP4 home page http://www.ccp4.ac.uk ***
Taiana Oliveira wrote:
Hi everybody,
I would like for you to clarify me abou the I/sigma value. I found
articles saying that resolutions with value below 2.0 must be
discarded but that for maxium likelihood methods this rule may be
ignored. How to proceed? What values may be considered as limits
in this case?
In my experience a value of 2.0 is conservative and seems to be
used mostly to avoid arguments with reviewers. There is not a hard
cutoff value but I think you're likely ok to 1.5. As the I/SigI
goes down the information content gets lower but maximum likelihood
shoult account for that. Check Rfree in the highest resolution
shell to make sure there is at least some correlation between model
and observations.
Bart
==============================================================================
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
==============================================================================
begin:vcard
fn:Eleanor Dodson
n:Dodson;Eleanor
email;internet:[EMAIL PROTECTED]
tel;work:+44 (0) 1904 328259
tel;fax:+44 (0) 1904 328266
tel;home:+44 (0) 1904 424449
version:2.1
end:vcard
