A paper by a crystallographer fired from his university for misconduct has been partially retracted.
Last year, we covered the case of Robert Schwarzenbacher, formerly of Salzburg University. Schwarzenbacher had provided the crystallographic data for a paper in the Journal of Immunology, but those results raised questions with another crystallographer and prompted an investigation by the university. Schwarzenbacher admitted he’d committed misconduct, although he recanted at one point, and was eventually fired.
Now, the authors have retracted the crystallographic data from the Journal of Immunology paper. Here’s the partial retraction, which is listed as a correction:
Zaborsky, N., M. Brunner, M. Wallner, M. Himly, T. Karl, R. Schwarzenbacher, F. Ferreira, and G. Achatz. 2010. Antigen aggregation decides the fate of the allergic immune response. J. Immunol. 184: 725–735.
The authors wish to retract the crystallographic section of this article, in particular Fig. 8A (ribbon diagram of the crystal structure of Bet v 1d) and Fig. 8B (electrostatic surface potential map of Bet v 1a, left, and Bet v 1d, right).
An analysis published in Acta Crystallographica Section F. 68: 4, p. 366–376 resulted in an investigation by the Austrian Agency for Research Integrity. The Austrian Agency for Research Integrity has confirmed that Robert Schwarzenbacher concedes he committed scientific misconduct concerning the crystallographic data in this article. Furthermore, the Austrian Agency for Research Integrity confirms that the other data in the article are valid and that there is no suspicion of scientific misconduct apart from that in Fig. 8A and Fig. 8B.
The Protein Data Bank has been informed and has retracted the 3K78 entry.
Retraction of the crystallographic section does not affect the major conclusions of the article.
The paper has been cited 18 times, according to Thomson Scientific’s Web of Knowledge.
We asked the Austrian Agency for Research Integrity for a copy of their investigational report, but they said their rules only allowed them to release the report to the university and the researchers involved. Salzburg has not responded to our requests for the report
I just want to cite the “Guidelines for retracting articles” issued by the Committee on Publication Ethics:
“The term corrigendum (or correction) usually refers to an author error.[)] Partial retractions are not helpful
because they make it difficult for readers to determine the status of the article and which parts may be relied upon.”
I suppose that parts “which may be relied upon” are all those parts which were not retracted, assuming that materials and conclusions in the surviving article are not (too) damaged by the partial retraction. This is probably the reason why the notice in the Journal of Immunology reads “Retraction of the crystallographic section does not affect the major conclusions of the article”.
Although one could ask: If the X-ray structure was unimportant for the conclusions you draw, why did you include the structure in your article?
The last question you aks sums it up. If I retract parts of a paper due to misconduct, how can I expect the readers of my journal to have any confidence at all in the rest? Just because I claim that the misconduct does “not affect the major conclusions of the article”?
With this justification, I could just send in fraudulent papers. Even though I fabricated the data, there is a chance that my “major conclusions” are right, despite never having done any experiment. Then, the journal just retracts some experimental section, but the conclusions are still right… nice.
That is, we can now all commit misconduct, we just have to draw the right conclusions to not get the paper retracted (just parts of it).
Furthermore, I am rather puzzled by the wording of the notice: “The authors wish to retract…” – does that mean that the editor-in-chief or the editorial board did not want to retract anything? Or were they just not aware of the investigation that went on in Austria?
“The Austrian Agency for Research Integrity has confirmed that Robert Schwarzenbacher concedes he committed scientific misconduct concerning the crystallographic data in this article.”
My understanding is Robert Schwarzenbacker conceded that the submission he made to the protein databank contained errors, but denied misconduct. Last I heard his appeal against his dismissal was on-going, with the legal form of his dismissal being held as valid, but whether it was a justified dismissal not yet decided
http://www.meinbundeslandheute.at/salzburg/2012/11/06/CMS1352191971700/
Has the Austrian Agency for Research Integrity actually done a formal investigation into what actually happened, by which I mean produced a report? Did he actually falsify everything from scratch or did he construct crystals, collect data and then improved his structure from related proteins (if that is possible).
It always amazes me how the scientific community moves heaven and earth to protect some people accused of misconduct and dispatches others without even a facade of due process.
It says above that they did a formal investigation, they’re just not sharing the results. Regardless, an earlier informal investigation (by an independent crystallographer) showed that it would be difficult to arrive at this result by accident. The lengthy (open-access!) write-up is here:
http://journals.iucr.org/f/issues/2012/04/00/wd5176/index.html
There is nothing wrong with improving the *atomic model* using related proteins – this is standard practice. The problem here is that the underlying diffraction data simply couldn’t have been obtained as described, and show every sign of being faked – for instance, the data are obviously generated from a theoretical model, yet somehow have experimentally measured errors… To put it in terms more familiar to this audience, it’s basically the crystallographic equivalent of drawing an entire gel image in Photoshop and using the blur filter and smudge tool to make it look realistically messy.
Perhaps there is an innocent explanation, but none has been forthcoming, and it’s really difficult to come up with a “result” like this by accident. Also note that if this really was an accident (which I find frankly incredible), Schwarzenbacher had every opportunity to present the genuine data. Unless he had some unusually pathological data, the re-analysis should not have taken more than a couple of days for someone with his experience.
Regarding the reference mentioned in the previous post, don’t miss the response issued one year ago:
http://journals.iucr.org/f/issues/2012/04/00/wd5180/wd5180.pdf
The puzzling sequence is “Author Schwarzenbacher admits to the allegations of data fabrication…” followed by a note added in proof: “subsequent to the acceptance of this article for publication, author Schwarzenbacher withdrew his admission of the allegations.”
Well, I only quickly skimed throgh the paper, but it does look like the X-ray was really not that important. They discuss that the Bet v 1d (the pollen protein that was supposedly analysed) was structurally very similar to the previously known Bet v 1a structure. Slight differences in surface charge the authors found unlikely to change Antibody affinity/recognition, something the also showed in immunological experiments. And they show that Bet v 1d shows self-aggregation because of a disulfide link, all of which is also experimentally shown, and they just discuss that because of the X-ray-structure with the involved Cysteine residue hidden, this has to happen with Bet v 1d at least partly unfolded. That’s pretty much it as far as I can see in a few minutes reading. Apparently, no novel conclusion was drawn because of the crystallography, it only seemed to support a few findings in the immunologic experiments.
So the crystallography was used to make the paper more sexy. But it does not seem to be relevant to the paper at all. Seems like Schwarzenbacher did not want to go through all the work . Perhaps he figured nobody would look too close, since the highly related Bet v 1a was already available and the structural changes seemded quite predictable. In my opinion, even for a J Immunol paper, a simple in-silico derivation of the Bet v 1d structure using the known structure of the Bet v 1a protein should have been sufficient for everybody, including the journal. This is quite a common procedure and after all, the immunologic experiments in the paper are what’s important.
While I generally prefer clear cut retractions, I can see the issue here. There is a whole body of good and important work. And there is a very special type of work that hardly anyone understands in detail and where co-authors have to pretty much trust the author who did (or pretended to do) the work.
You could say they should have left the crystallographic work out because you should not be co-author of something you don’t understand. But that would mean there could not be any interdisciplinary work with crystallography, since basically nobody else but the crystallographers truely understands what they are doing.
And what should be done with the sound data from retracted papers? This has not really been adressed by the “retraction community”. Journals seem highly unwilling to republish sound portions of retracted papers. Personally, I would even cite a retracted paper, if I would replicate an experiment from such a paper, especially if it was an obviously sound part of that paper.
I also note that the notice released by the Journal of immunology starts with “The authors wish to retract the crystallographic section of this article…”. This is obviously impossible, because one author, Professor Gernot Achatz passed away on 14 August 2011, two years before the partial retraction. I’m afraid he never signed that notice.