PNAS retraction marks second for crystallography group

Two crystallographers who retracted a Structure paper last year have retracted a study about a similar subject in the Proceedings of the National Academy of Sciences, for similar reasons.

Here’s the notice for the paper, which has been cited 23 times, according to Thomson Scientific’s Web of Knowledge:

Retraction for “Structural basis for nucleotide exchange on Gαi subunits and receptor coupling specificity,” by Christopher A. Johnston and David P. Siderovski, which appeared in issue 6, February 6, 2007, of Proc Natl Acad Sci USA (104:2001–2006; first published January 30, 2007; 10.1073/pnas.0608599104).

The authors wish to note the following: “In our paper, a cocrystal structure at 2.2 Å resolution was described of the heterotrimeric G-protein alpha subunit Gαi1 bound to two peptides: one from an artificial sequence that promotes nucleotide exchange (KB-752) and a second peptide (D2N) from the third intracellular loop of the D2 dopamine receptor (PDB ID code 2HLB). Further examination of the unbiased electron density map has revealed that, while electron density exists for the KB-752 peptide, there is a lack of clear and continuous electron density for the D2N receptor peptide in the complex. Because the structural model represents a major conclusion of the paper but is unsupported by the experimental electron density map, we wish to retract the paper. Both authors deeply regret this mistake and sincerely apologize.”

The reason for the retraction is essentially the same as the one for the Structure paper. We’ve contacted Siderovski for comment, and will update with anything we hear back.

Siderovski has collaborated with the late Maria Diverse-Pierluissi, four of whose papers were retracted after her death. Siderovski and Diverse-Pierluissi published two papers together, neither of which has been retracted — although a former postdoc in Diverse-Pierluissi’s lab suggests there may be issues with one of them. And Mount Sinai, where Diverse-Pierluissi worked, told us several months ago that they considered their investigation closed.

In other words, we have no reason to think the retractions are related. But this seems to be the first case of two groups linked by disparate retractions.

Update, 2 p.m. Eastern, 1/26/12: It turns out we missed a correction by the same group that was released the same day in PNAS — and for basically the same reason as the retraction:

Correction for “Regulators of G-protein Signaling accelerate GPCR signaling kinetics and govern sensitivity solely by accelerating GTPase activity,” by Nevin A. Lambert, Christopher A. Johnston, Steven D. Cappell, Sudhakiranmayi Kuravi, Adam J. Kimple, Francis S. Willard, andDavid P. Siderovski, which appeared in issue 15, April 13, 2010, of Proc Natl Acad Sci USA (107:7066–7071; first published March 29, 2010; 10.1073/pnas.0912934107).

The authors note that Fig. 1 appeared incorrectly. They also wish to note the following: “Upon further evaluation of the unbiased electron density maps for the structural models of Gαi1(G202A)·GDP (PDB id 2PZ2) and Gαi1(G202A)·GDP·AlF4–(PDB id 2PZ3), there is a lack of clear and continuous density to support an entirely ordered switch II region or to support the presence of aluminum tetrafluoride in the latter structure. We have therefore obsoleted the x-ray structure model PDB 2PZ3 from the Protein Data Bank. We have replaced PDB 2PZ2 with PDB 3UMS in the Protein Data Bank to reflect the more accurate refinement of the Gαi1(G202A)·GDP structural model.”

The corrected figure and its corresponding legend appear below. This error does not affect the conclusions of the article.

Hat tips: Clare Francis, Vladimir Svetlov

22 thoughts on “PNAS retraction marks second for crystallography group”

  1. I would check more deeply into any possible connection with M. Diverse-Pierluissi as it appears that the majority of her work was tainted…there may be more. We still don’t know how she died; I hate to be rude, but it does have some potential significance. Any more detail on these connections, Maestro Oransky?

    1. The TIBs paper published together by these two incorrectly suggests that GoLoco was a G protein GEF, rather than a GDI as it turned out to be.

      The GoLoco motif: a Galphai/o binding motif and potential guanine-nucleotide exchange factor.
      Siderovski DP, Diversé-Pierluissi M, De Vries L

    2. Thanks for the question. We’ve reported everything we know about this case, but will continue to update as we learn more.

  2. Cases like this are why the field of crystallography is fortunate to have a resource like the protein data bank (PDB) where authors must deposit their coordinates and structure factors. However, I wish that there could be some consensus about submitting maps and coordinates to reviewers before publication. Since the peptide is the novel part of the paper in question, it would have been the first thing the reviewers looked at, and any competent structural biologist would have caught this problem.

    1. It seems hard to believe that a competent structural biologist was among the reviewers, or if they were that the peptide was the first thing they looked at as in the case of the two retracted papers figures showing the maps these peptides were built into are (IMHO) enough to call into doubt the validity of the models:

      1. Also of interest: try recalculating the maps with the deposited model (2HLB) and structure factors, and look at the 2Fo-Fc map with the same contour level reported in Figure S5.

      2. It was published in PNAS. The editor, Lefkowitz, and Siderovski do have at least one joint publication. It’s possible that Lefkowitz was a pre-arranged editor, so the paper may not have undergone the usual rigors of peer review. You’ve gotta love PNAS and its multiple ways to publish sketchy stuff.

      3. Spot on sfs. PNAS has an interesting reviewing process!! A member can publish certain number of papers per year and he can edit as many as possible. Prearranged editor is a nice thing – as collaborators (current or former), former colleagues, former postdocs can always request a member to be the editor of their paper without members having to sacrifice their quota per year…how about that..I am just stating the fact and this is not a complaint…

    2. Re: Anonymous crystallographer

      Looking at the maps and coordinates in COOT, not only is there no density for the peptide (and b-factors range from 150-180), but the peptide contains 2 cis bonds, despite having no prolines.

    3. We are also fortunate to have such resources as PDBREPORT and PDB-REDO, because frankly the consumers of the structural information (like myself) far outnumber the people who as suggested below, could “recalculate” the maps with the models, etc. Unless there is a good reason usually even people who know the craft don’t recalculate each other structures. Personally I believe it is of essense to raise the computational and crystallographic literacy outside the structural field, among the folks who treat “structures” (as in pdb) as physical reality and try to do modeling and make mutants based on trash.

  3. I think it’s worth mentioning that the validation summary for referees proposed by the Validation Task Force (see for more details) would have clearly highlighted problems with the D2N receptor peptide.

    If the submission of this summary with the manuscript describing the structure becomes mandatory it might be hoped that problematic structures would be noticed during the review process.

  4. I feel this blog pays too much attention to crystallographers and crystallography
    ( a field characterized by a profound lack of creativity and which I find exquisitively boring)

    1. I find this thread quite useful. As a consumer (but not a producer) of structural data, insight into whether the presented models are valid is appreciated. The field is sufficiently complex that most of us on the outside rely almost completely on the journal and reviewers to verify the validity of the model. So the questions raised here are quite interesting. From what I gather, the crystallographers are implying that the papers might not have been reviewed by crystallographers? It seems like a lot of money and time could be saved if there were established standards for reviewing these data (as mentioned above). There are rumors that at least one more paper from the group will be corrected/retracted.

    2. Much of what we know of biomolecular mechanism and the underpinning of rational drug design (two give just two examples) comes from biomolecular crystallography. It’s a massively important subject and has the advantage of giving hard and essentially incontrovertable information (so long as it’s done properly). It’s a field with rather little fraud in my experience but one in which efforts to ensure quality control is paramount.

      I wonder whether your regrettable boredom might be encapsulated in the saying re mountaineers and specifically those that eschew certain mountains which are considered (by some) to be “boring” viz:

      “There are no boring mountains, only boring mountaineers”

      Happily the Nobel Prize committees didn’t find the insightful and productive crystallographic determination of potassium ion channel structures by Rod McKinnon or the astonishing structures of the ribosome by Ramakrishnan, Steitz and Yonath (two name two recent examples) to be “characterized by a profound lack of creativity” or “exquisitively boring”.

      1. @chris: Your argument lacks creativity
        and it’s is exquisitevly boring. I presume you are a crystallographer.
        By the way, the McKinnon and RSY
        work were above average advances for the respective fields but in my opinion not Nobel prize worthy discoveries

      2. Well it’s a shame that you are so easily bored MT Orr. Bad luck!

        No I’n not a crystallographer (molecular biologist), ‘though I have found MacKinnon’s work to be hugely impactful on my own research. You must have astonishingly high standards not to consider a full atomic resolution structure of the ribosome not to be a remarkable achievement and worthy of a Nobel Prize.

        Likewise RodMackinnon’s ion channel structures. The 7 structures he published between 1998 and 2004 completely revolutionised the field of membrane proteins and brought a number of long-studie elements of ion channel phenomena to sharp focus (e.g. the nature of voltage and ligand gating; the mechanism of voltage sensing; the molecular basis of ion selectivity) and have formed much of the basis of rational pharmaceutical drug development in the ion channel field. He provided a huge impetus to the successful efforts to crystalize membrane proteins. The entire field of drug development and screening for unwanted side effects arising from adventitious interactions with heart ion channels is now pursued in the knowledge of the structural basis of these effects.

        MacKinnon’s ion channel crystal strucutre papers have been cited around 1000 times each on average (the original KcsA structure paper has over 3500 citations).

        In other words a revolutionary piece (or pieces!) of work that has had by far the greatest impact on nerve transmission and ion channel research in the last two decades. Certainly deserving of a Nobel Prize by anyone’s standards….

        ..except yours apparently…

      3. I think we’re being trolled – this is only the third post that mentions crystallography, in contrast to dozens that mention Western blots. (For what it’s worth, I find Western blots boring and unimaginative, but those do seem to get retracted frequently, so their presence on this blog is entirely appropriate.)

  5. I have used the Siderovski Structure retraction as a case study for teaching my group about scientific dishonesty. I was getting ready to give this presentation to another crystallography audience at my institution tomorrow when I discovered this site informing me that there is a second retraction from this group.

    I found another dubious result from this group. I viewed the electron density (from the EDS server) from another of Siderovski’s low resolution structures, 2XNS. The paper describing the structure was published in a very good journal, JACS, in 2011, v133, p. 4190. The structure is a mess. The resolution is 3.4 Angstrom, yet they include water molecules and model alternative conformations for some side chains. The B-factors are well beyond 100. If anyone is interested in investigating this structure further, check out the density for His513 of chains C and D.

    It is an outrage that Siderovski still has NIH funding.

    1. “Scientific dishonesty” is one explanation for retraction, “scientific ignorance” is also a distinct possibility. Siderovski does not have a formal crystallographer’s training to my knowledge and may have simply failed (inexcusably) to provide oversight to a misguided graduate student. I have heard rumor of an institutional and independent investigation into these matters at Siderovski’s former home of UNC Chapel Hill, which did not find evidence to support a charge of academic dishonesty.
      Regarding 2XNS, I too have investigated that publication. The structure has good geometry and reasonable density quality given the resolution; B-factors are a poor judge of quality at low resolution. There are, count them, 1…2 water molecules, and yes the alternative side chains are a stretch, but not impossible, at 3.4 Angstrom. But if one looks at the conclusions of the JACS paper, they are well-supported by what I see in the structure. The simple questions is: is there an alpha helix at a particular location that looks like the Rosetta design? The answer is unequivocally “yes”, given these low resolution data. The resolution limitations of the structure are also acknowledged directly.
      Let’s not get out the pitchforks and torches over two water molecules and bad side chains, and remember that a structure, like any other experimental data, only needs to support conclusions as advertised in a paper.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.