I must add something...
ID14-2 beam line in our hands produced during first decade of 2000's
data sets that
for structures at about 1.8 - 1.6 Angstrom constantly leaded to a very
good R - Rfree (in the range of 12 %-18%)
As an example see PDB entry 1Y9A. If will be needed I will supply
diffraction data for several structures. Such quality produces
problems as a referee
accused us by "over-fitting Rfree" whatever it means. He also "new"
what is a theoretical difference of R-Rfree in any resolution?
During our discussions recently with Dr Alexander Popov, responsible
for the ESRF ID23-1 beam line on this matter Dr Popov suggest that
it is
because the X-ray beam on this station is very large. I also think
that its is largely parallel, old quality of the synchrotron lines.
Maybe other people who noticed similar thing may add more.
BTW final data set scaling statistics (SCALEPAK) is shown:
Shell Lower Upper Average
Average Norm.
Linear Square
limit Angstrom I
error stat. Chi**2 R-
fac R-fac
20.00 3.79 7081.7 258.9 136.6 1.172
0.029 0.031
3.79 3.01 3894.0 157.2
95.2 1.278 0.038 0.040
3.01 2.63 1514.9 97.1
64.8 0.956 0.056 0.140
2.63 2.39 938.8 75.4
57.5 0.976 0.074 0.080
2.39 2.22 717.1 74.5
59.5 0.925 0.093 0.100
2.22 2.09 534.2 71.6
62.2 0.947 0.122 0.116
2.09 1.99 398.6 70.6
65.1 0.915 0.164 0.169
1.99 1.90 258.1
69.1 66.3 0.957
0.256 0.299
All reflections 1875.3
108.1 75.5 1.018
0.051 0.046
Dr Felix Frolow
Professor of Structural Biology and Biotechnology
Department of Molecular Microbiology
and Biotechnology
Tel Aviv University 69978, Israel
Acta Crystallographica D, co-editor
e-mail: mbfro...@post.tau.ac.il
Tel: ++972 3640 8723
Fax: ++972 3640 9407
Cellular: ++972 547 459 608
On Mar 20, 2009, at 2:29 PM, James Holton wrote:
Yes, Harry, indeed there is a program for simulating diffraction
patterns. You can get a development snapshot of it here:
http://bl831.als.lbl.gov/~jamesh/mlfsom/development_snapshot.tar.gz
MLFSOM (mosflm in reverse) is not the only program of its kind in
existence and I don't think it is a good idea to keep the fact that
they exist a secret in any way. In fact, MLFSOM has been extremely
instructive in establishing how important different sources of error
are to the structure determination process, and where the "threshold
of solvability" is in real-world units. I am writing this up now
and I have given several talks about it recently but it has always
puzzled me that noone has EVER asked me if there is some way to
"identify as fake" the images that come from MLFSOM. The answer to
this question is: "Yes, there are several".
Moving on...
When it comes to Garib's original question of "do we need the
images", I am definitely of the opinion that the answer to this
question is "yes". In fact, I would like to ask Garib and everyone
else if they can answer the question: Why is Rcryst/Rfree from
almost every protein crystal structure on the order of 20% when the
intensities are generally measured to better than 5%? What are we
missing? Small molecule structures are unpublishable with Rcryst >
Rsym because this means that the presented model does not explain
the observations to within experimental error. I will tell you for
nothing that if you feed fake data from MLFSOM into Elves and ARP/
wARP, you will get back an Rcryst/Rfree that is roughly equal to
Rsym (~5-6%), so this means that none of the sources of error I have
included in MLFSOM: photon-counting noise, shutter jitter, beam
flicker, sample vibration, diffuse scatter, absorption effects,
funny spot shapes, radiation damage, detector read-out and
calibration noise. None of these sources of error are big enough to
explain the "R-factor gap" in macromolecular crystallography.
So, if you don't know what it is we are missing, how can you be sure
it is not in the image data?
-James Holton
MAD Scientist
harry powell wrote:
Hi
I've heard of a tool from the Golden State which could
(potentially) be used for forging diffraction images... I believe
it's called "mlfsom".
On 18 Mar 2009, at 17:50, Felix Frolow wrote:
One convincing argument I have:
We will be able to catch fraud ultimately. Fraud is a devastation
for structural biology.
...Unless they will be smart enough to forge diffraction data
images, not a big deal.
The second one - in the case of a controversy of the deposited
results (possible thing) we can try to re-interpret the space
group and Bravais lattice
And one more, when we have time we can show that we know better to
process and to refine ;-)
Dr Felix Frolow
Professor of Structural Biology and Biotechnology
Department of Molecular Microbiology
and Biotechnology
Tel Aviv University 69978, Israel
Acta Crystallographica D, co-editor
e-mail: mbfro...@post.tau.ac.il
Tel: ++972 3640 8723
Fax: ++972 3640 9407
Cellular: ++972 547 459 608
On Mar 18, 2009, at 6:41 PM, Garib Murshudov wrote:
Dear all
Before going into and trying to find a technical solution to the
problem it would be good if decide if we need images. As far as I
know if we face with a problem to solve and we know that it is
necessary to solve then we find technical solution to the problem
(either from other fields or we find our own solution with some
elements of reinvention of new MX wheels).
Do we need images to store? What kind of information we can
extract from images that we cannot from amplitudes, intensities
(even unmerged)? Does anybody have a convincing argument for
favour of images?
regards
Garib
On 18 Mar 2009, at 16:32, Herbert J. Bernstein wrote:
Actually the radiologists who manage CT and PET scans of brains
do have
a solution, called DICOM, see http://medical.nema.org/. If we
work
together as a community we should be able to do as well as the
rocket scientists and the brain surgeons' radiologists, perhaps
even
better. -- Herbert
=====================================================
Herbert J. Bernstein, Professor of Computer Science
Dowling College, Kramer Science Center, KSC 121
Idle Hour Blvd, Oakdale, NY, 11769
+1-631-244-3035
y...@dowling.edu
=====================================================
On Wed, 18 Mar 2009, Jacob Keller wrote:
Apparently it DOES take a rocket scientist to solve this
problem. Maybe the brain surgeons also have a solution?
JPK
*******************************************
Jacob Pearson Keller
Northwestern University
Medical Scientist Training Program
Dallos Laboratory
F. Searle 1-240
2240 Campus Drive
Evanston IL 60208
lab: 847.491.2438
cel: 773.608.9185
email: j-kell...@northwestern.edu
*******************************************
----- Original Message ----- From: "Klaas Decanniere" <klaas.decanni...@vub.ac.be
>
To: <CCP4BB@JISCMAIL.AC.UK>
Sent: Wednesday, March 18, 2009 5:36 AM
Subject: Re: [ccp4bb] images
Herbert J. Bernstein wrote:
Other sciences have struggled with this and seem to have found
an answer.
Have e.g. a look at http://heasarc.nasa.gov/docs/heasarc/fits.html
kind regards,
Klaas
This is a good time to start a major crystallogrpahic image
archiving effort. Money may well be available now that will
not be
avialable six month from now, and we have good, if not perfect,
solutions available for many, if not all, of the technical
issues
involved. Is it really wise to let this opportunity pass us
by?
The deposition of images would be possible providing some
consistent
imagecif format was agreed.
This would of course be of great use to developers for
certain
pathological cases, but not I suspect much value to the user
community - I down load structure factors all the time for
test
purposes but I probably would not bother to go through the
data
processing, and unless there were extensive notes
associated with
each set of images I suspect it would be hard to reproduce
sensible
results.
Harry
--
Dr Harry Powell, MRC Laboratory of Molecular Biology, MRC Centre,
Hills
Road, Cambridge, CB2 0QH
