Yes, science is done by humans and mistakes will be made. Kudos
to the lab involved for owning up the mistake and doing the right
thing. It takes integrity to stand up in public and say "I screwed
up".
However, as a community of artisans we should not pass up the
opportunity to learn from these failures. When an airplane crashes
there is a detailed investigation because, as a society, we don't like
planes falling from the sky. When there is a problem like this we
should look into it and see what can be learned.
From what I've seen, the lessons here are basically
1) The crystallographer didn't have the training to distinguish between
"the refinement program ran" and "the model is correct." There was
a failure to apply known validation criteria and the failure of
this model to meet community-wide standards was unrecognized.
2) The crystallographer didn't have the training to understand how
to distinguish between a ligand binding and one not binding.
3) During the creation of Table 1 it was not noticed that the stats
presented did not match the model. There was some sort of failure
in communication among the authors.
4) During deposition the processing entity (This was the RCSB)
did not recognize that the deposited rmsds did not match the
model. While the bond length rmsd was listed as 0.010 A the
list of deviants greater than 6 sigma was many hundreds of
entries long. This contradiction should have been noted and
and explanation demanded of the depositor.
Lesson #4 can probably be quickly applied by a change in the
process used by the annotators. I don't expect the wwPDB to be
able to exactly match the reported rmsd's because people use
differing libraries and use differing weighting, but they should
not be this far off!
Lessons #1 - 3 are harder to define without access to the
lab records and "post-crash" interviews. We, on this BB, know
the solution to lack of training, but I expect that the people
who don't provide the required training to their students don't
follow this BB. I suspect they don't read Letters to the Editor
in acta D either.
I don't know how to get the word out to those biochemistry
P.I.'s who mentor students using crystallographic techniques that
just pushing the buttons on user-friendly software is not good
enough. This is the question facing the crystallographic
community that desperately needs an answer.
Dale Tronrud
On 08/11/11 13:33, Maia Cherney wrote:
> As the macromolecular crystallography becomes more automated and
> user-friendly many biologists learn to solve structures and they can
> make mistakes. Besides, new data become available that can give new
> ideas etc. I don't think that's so horrible to make an honest mistake
> and retract papers. Even great scientists sometimes published papers
> with wrong analysis.
>
> Maia
>
> On 11/08/2011 1:05 PM, Judith Murray-Rust wrote:
>> Just another point - the macromolecular community are not the only
>> ones with a problem - I've just been shown
>> http://retractionwatch.wordpress.com/
>> which sheds some light on retractions. And also maybe says something
>> about why original data should be available/part of the review process.
>> J
>> ________________________________________
>> From: CCP4 bulletin board [[log in to unmask]] On Behalf Of Gloria
>> Borgstahl [[log in to unmask]]
>> Sent: 11 August 2011 19:32
>> To: [log in to unmask]
>> Subject: Re: [ccp4bb] Another paper& structure retracted
>>
>> Dale, This is exactly the conversation I just had with my student
>> Jason, right on!
>> The paper we are writing just now, this is figure 1.
>> But I always get rejected by Nature, so go figure.
>> On Thu, Aug 11, 2011 at 1:25 PM, Dale
>> Tronrud<[log in to unmask]<mailto:[log in to unmask]>>
>> wrote:
>> I agree with Prof. Tomchick: if the point of your paper is your
>> crystal structure of
>> the binding of a ligand to a protein you should include a figure with
>> the omit map
>> (displayed without a "cover radius") that convinced you that binding
>> took place. I
>> prefer that map over some simulated, after-the-fact, omit map
>> calculated just for
>> publication.
>>
>> This is not simply a matter for reviewers to be gatekeepers, it is
>> important for the
>> readers to know what level of confidence to place in this result, and
>> it is instructional
>> for everyone to see what ligand binding density looks like.
>> Apparently some people don't
>> know what features to look for to distinguish between signal and noise.
>>
>> Dale Tronrud
>>
>> On 08/11/11 09:40, Diana Tomchick wrote:
>>> A quick glance at the header of the PDB file shows that there is one
>>> glaring discrepancy between it and the table in the paper that hasn't
>>> been mentioned yet in this forum. The data completeness (for data
>>> collection) reported in the paper is 95.7%, but in the header of the
>>> PDB file (actually, in both the 2QNS and the 3KJ5 depositions) the
>>> data completeness (for data collection) is reported as only 59.4%.
>>> The PDB header also contains an inconsistency, with the data
>>> completeness (for refinement) reported as 95.7%. Since the numbers of
>>> reflections reported for refinement versus data collection in the PDB
>>> header differ by less than 1%, it appears that there's been a bit of
>>> magical thinking that took place somewhere along the process from
>>> data processing to final model refinement. Small wonder that the
>>> refined geometry is so poor. Perhaps if these scientists had actually
>>> collected a complete dataset, we would not be having this conversation.
>>>
>>> Diana
>>>
>>> P.S. I have, on occasion, provided the coordinates and a map file to
>>> reviewers when they requested it. The last time it was requested was
>>> many years ago; I decided it was safer and easier if I provided as
>>> much information as possible in the manuscript (including better
>>> quality electron density figures than appear in this paper) to allow
>>> the reader to determine whether the work is valid or not.
>>>
>>> * * * * * * * * * * * * * * * * * * * * * * * * * * * *
>>> Diana R. Tomchick
>>> Associate Professor
>>> University of Texas Southwestern Medical Center
>>> Department of Biochemistry
>>> 5323 Harry Hines Blvd.
>>> Rm. ND10.214B
>>> Dallas, TX 75390-8816, U.S.A.
>>> Email: [log in to unmask]
>>> 214-645-6383<tel:214-645-6383> (phone)
>>> 214-645-6353<tel:214-645-6353> (fax)
>>>
>>>
>>>
>>> On Aug 10, 2011, at 5:45 PM, Dale Tronrud wrote:
>>>
>>>> I've made a quick look at the model and the paper - and it doesn't
>>>> need more than a quick look. The description of the model in
>>>> the paper sounds great. The problems in the model are clear. My
>>>> favorite is the quote "Trp-477 of PTH1R makes several van der Waals
>>>> contacts with Trp-339 and Lys-337 of G-beta-1 ...". They are
>>>> "contacts"
>>>> all right. The distances between the 477:CH2 and 337:CE is 2.75 A
>>>> and between 477:NE1 and 339:CH2 is 2.26 A. There are many more.
>>>>
>>>> In general the geometry of this entire model is terrible. In
>>>> Table 1 the bond length rmsd is listed at 1.64 A and the bond angles
>>>> are 0.0078 deg! Perhaps one is to presume the numbers should be
>>>> swapped. In any case, the values I calculate for the model are
>>>> 0.160 A and 4.46 deg! Absolutely dreadful. The PDB header lists
>>>> the (swapped) values from the paper and then reports hundreds of
>>>> outliers.
>>>>
>>>> The tools proposed by the Validation Task Force should cause a
>>>> model like this to pop out clearly. Even the old tools show this
>>>> model is quite unreliable. We just have to use them.
>>>>
>>>> Dale Tronrud
>>>>
>>>> On 08/10/11 14:35, Jacob Keller wrote:
>>>>> On the surface it doesn't seem as bad as others, i.e., it does not
>>>>> seem to be a real fake--perhaps just a strong form of wishful thinking
>>>>> and creative density interpretation. I wonder what would be a good
>>>>> metric in which to establish a cutoff for present/not present in
>>>>> density. CC, maybe?
>>>>>
>>>>> Jacob
>>>>>
>>>>> On Wed, Aug 10, 2011 at 4:01 PM, David
>>>>> Schuller<[log in to unmask]<mailto:[log in to unmask]>> wrote:
>>>>>> Time to fuel up the gossip engines for the approaching weekend:
>>>>>>
>>>>>>
>>>>>> http://www.sciencedirect.com/science/article/pii/S096921260800186X
>>>>>>
>>>>>> RETRACTED: Structure of the Parathyroid Hormone Receptor C
>>>>>> Terminus Bound to
>>>>>> the G-Protein Dimer G
>>>>>> Structure, Volume 16, Issue 7, 9 July 2008, Pages 1086-1094
>>>>>> Structure 2QNS withdrawn.
>>>>>>
>>>>>> --
>>>>>> =======================================================================
>>>>>>
>>>>>> All Things Serve the Beam
>>>>>> =======================================================================
>>>>>>
>>>>>> David J. Schuller
>>>>>> modern man in a post-modern world
>>>>>> MacCHESS, Cornell University
>>>>>>
>>>>>> [log in to unmask]<mailto:[log in to unmask]>
>>>>>>
>>>>>
>>>>>
>>>
>>> ________________________________
>>>
>>> UT Southwestern Medical Center
>>> The future of medicine, today.
>>
>> NOTICE AND DISCLAIMER
>> This e-mail (including any attachments) is intended for the
>> above-named person(s). If you are not the intended recipient, notify
>> the sender immediately, delete this email from your system and do not
>> disclose or use for any purpose.
>>
>> We may monitor all incoming and outgoing emails in line with current
>> legislation. We have taken steps to ensure that this email and
>> attachments are free from any virus, but it remains your
>> responsibility to ensure that viruses do not adversely affect you.
>> Cancer Research UK
>> Registered in England and Wales
>> Company Registered Number: 4325234.
>> Registered Charity Number: 1089464 and Scotland SC041666
>> Registered Office Address: Angel Building, 407 St John Street, London
>> EC1V 4AD.
>>
|