Software glitch — not intentional manipulation — sunk immunology paper, says author

kuo photo
A black box appears over the control lane on the left

New evidence suggests a retracted paper was felled not by intentional manipulation — as it first appeared — but by a software glitch.

In 2014, we reported that Biochemical Journal had retracted a paper on suspicion it contained “shoddy Photoshopping”  — someone appeared to have blacked out a control lane in one figure. Now there’s evidence that it wasn’t done on purpose: An investigation at Duke into eight papers, including the Biochemical Journal paper, did not find evidence of misconduct; lead author Paul Kuo, currently chair of surgery at Loyola Medicine, told us that a glitch in the software caused the black box. Nevertheless, the journal does not plan to un-retract the paper.

Kuo has one other retraction, and a number of his papers have been questioned on PubPeer.

He set the record straight about the 2014 retraction in a recent comment on the PubMed entry for the paper:

At the time of the retraction, the original data were no longer available and we were unclear as to the reason for the “black box”. Subsequently, the following has occurred: 1. The study has been repeated using the original samples and the original findings were replicated, 2. Dr. Guo discovered an ImageJ software artifact or glitch that produced the “black box”. The glitch has been submitted to both ImageJ for correction and to an imaging journal for publication. It is displayed in this video: https://www.dropbox.com/s/7sl0qh7woqxljvt/Backspace.wmv?dl=0  3. The original paper and our subsequent discoveries have been reviewed by the Duke Institutional Research Review Board; their findings were that the research record was not misrepresented and there was no evidence for research misconduct. 4. When the journal was informed of these, they declined to retract or correct the original retraction. Hence, the reason for this post.

Donna Cookmeyer, a Research Integrity Officer at Duke, was not able to speak to Retraction Watch directly about the university’s investigation. However, she confirmed that the following statement from Kuo’s lawyer was accurate:

At Dr. Kuo’s request, and with the approval of Duke University’s Office of Research Integrity and Office of General Counsel, Dr. Cookmeyer, Research Integrity Officer confirms that Duke’s thorough investigation of the Biochem J paper concluded there was no research misconduct and no misrepresentation of the scientific record’

The editor in chief of Biochemical Journal, David Carling, told us the journal has no plans to lift its retraction, despite Kuo’s explanation:

Whilst we appreciate this possibility, unfortunately, the authors were unable to provide the original image prior to software processing. The authors did send us a gel in which they had re-analysed the original samples, but the quality of this image was extremely poor and it was not possible to draw any firm conclusions from the re-analysis. After careful consideration, we concluded that without the original pre-processed image there were no grounds for changing the retraction notice.

Here’s the retraction notice for “Phosphorylation of Ser158 regulates inflammatory redox-dependent hepatocyte nuclear factor-4α transcriptional activity,” published when the source of the black box was unknown:

It has been brought to our attention that Figure 1(A) contains a background subtraction box in the lane labelled ‘Control + IgG’. As this band represented an important negative control, this has serious potential to affect the conclusions drawn in the paper, and therefore we are retracting this paper. We thank the Editorial Board of the Biochemical Journal for pointing out this error.

Kuo told us he wasn’t happy with how we covered his case in 2014:

[T]he original title and content of the Retraction Watch article and the associated comments have been extremely problematic for our research group. Luckily, the papers listed have been reviewed by institutional authorities and again, there is no evidence of research misconduct. However, when my name is Googled, this Retraction Watch article and its inflammatory (and defamatory?)  title come up on the second search page.

More of Kuo’s papers have been questioned on PubPeer. He told us:

Our lab group submitted detailed responses to all allegations from Pubpeer (and Clare Frances) related to 12 papers published during my time at Duke (2000-2010) and related to work performed at Duke but published after my departure. Subsequent to this, additional review (over the course of 1.5 years) in the form of face-to-face committee interviews with lab members, submission of lab notebooks and data from computer hard drives and imagers, repeat experiments using original samples, additional image analysis and in-person demos of software faults were conducted for selected papers. In total, the committee’s report involved 7 publications (in addition to the 2006 Biochem publication; all published between 2007 and 2012).

Ann Bradley, Associate University Counsel at Duke, confirmed that statement is accurate.

Kuo earned a second retraction last year for a Journal of Biological Chemistry paper on which he is one of six co-authors, “Differential expression of intracellular and secreted osteopontin isoforms by murine macrophages in response to Toll-like receptor agonists.” The retraction notice — in JBC’s vintage style — is pretty cryptic:

This article has been withdrawn by the authors.

The paper was published in 2010 and has been cited 36 times, according to Thomson Reuters Web of Science.

Kuo couldn’t tell us anything about that retraction; he said he just “reviewed the grammar and composition” for the paper. We have reached out to last author Chengjiang Gao, affiliated with Shandong University, for more information.

Update, May 20 1:50 PM:

We asked Wayne Rasband, the developer of ImageJ, for comment about Kuo’s explanation of what happened. He told us:

This glitch is due to a misunderstanding of how ImageJ works. Unlike some other programs, pressing the backspace key in ImageJ does not delete the selection box. Instead, it is a shortcut for the Edit>Clear command, which erases the contents of the selection box to the background color (usually black). I have attached the section of the ImageJ Users Guide that describes the Clear command. To prevent the selection box from being inadvertently cleared, the gel analyzing function in the latest version of ImageJ (1.51b10) disables this shortcut.

He sent us a screenshot of the explanation:

ImageJ

Like Retraction Watch? Consider making a tax-deductible contribution to support our growth. You can also follow us on Twitter, like us on Facebook, add us to your RSS reader, sign up on our homepage for an email every time there’s a new post, or subscribe to our new daily digest. Click here to review our Comments Policy. For a sneak peek at what we’re working on, click here.

 

 

23 thoughts on “Software glitch — not intentional manipulation — sunk immunology paper, says author”

  1. Um, when you submitted the paper and image files for publication, you didn’t pay attention to the black box?

  2. A couple of thoughts and questions…
    Why would you re-save the image file from a densitometry analysis? No need to save it at all since the densitometry doesn’t do anything to it. Why also, would you choose to use a post-densitometry saved image to prepare the images for publication?
    Also, the deleted black box in the paper in question is not a “pure” all black box, there are some gray pixels in it. How is the appearance of such a graded/pixelated box, consistent with the explanation in the video that appears to show a uniform black box is created?

  3. Have the programmers behind ImageJ weighed in with a comment? If their software is randomly blacking out bands, that is rather a big deal.

    1. I have been wondering the same thing. Finally sent them an email. Will post response.

      1. The developer of ImageJ, Wayne Rasband, replied to my email. The behavior shown in the video is not a bug and is in the ImageJ documentation. However, he does say: “To prevent the selection box from being inadvertently cleared when analyzing gels, the backspace key is ignored in the latest ImageJ daily build (1.51b10) if there is at least one outlined lane.” So, thanks to him for the response.

  4. ImageJ is a supreme program but like all software it is only as reliable as the user. More relevant – and as Brookes observes- Duke’s analysis (as described) would be more compelling, and more thorough, if they had simply asked whether other features in the evidence they did have (the questioned photo) were compatible with the explanation given in the video? Did the Duke investigators look for evidence of the residual margins of a missing band “outside of the box”. . . perhaps with 16 pixel high profile plot in ImageJ?

  5. J Pharmacol Exp Ther. 1997 Aug;282(2):1072-83.
    Nitric oxide and acetaminophen-mediated oxidative injury: modulation of interleukin-1-induced nitric oxide synthesis in cultured rat hepatocytes.
    Kuo PC1, Schroeder RA, Loscalzo J.
    Author information

    1Department of Surgery, University of Maryland Medical Systems, Baltimore 21201, USA.

    PMID: 9262377

    Figure 3A.
    http://i.imgur.com/YUDXcnv.jpg

  6. J Am Coll Surg. 2013 Jul;217(1):17-26; discussion 26. doi: 10.1016/j.jamcollsurg.2013.02.025. Epub 2013 Apr 23.
    Osteopontin up-regulates critical epithelial-mesenchymal transition transcription factors to induce an aggressive breast cancer phenotype.
    Li NY1, Weber CE, Mi Z, Wai PY, Cuevas BD, Kuo PC.
    Author information
    1Department of Surgery, Loyola University, Maywood, IL 60153, USA.

    2015 correction.
    http://www.sciencedirect.com/science/article/pii/S1072751515003427

    “In the article “Osteopontin Up-Regulates Critical Epithelial-Mesenchymal Transition Transcription Factors to Induce an Aggressive Breast Cancer Phenotype,” by Li and colleagues, which originally appeared in the July 2013 issue of the Journal of the American College of Surgeons, the beta-actin images in Fig 4 and Fig 7 were identical except that one is rotated 180 degrees. This was an inadvertent mistake. The results in the figure are valid, and the conclusions of the article are correct. The corrected and verified beta-actin images are presented here. The authors apologize for this error.”

  7. Dig Dis Sci. 2005 Jul;50(7):1288-98.
    Nitric oxide-dependent osteopontin expression induces metastatic behavior in HepG2 cells.
    Guo H1, Marroquin CE, Wai PY, Kuo PC.
    Author information

    1Department of Surgery, Duke University Medical Center, Durham, North Carolina 27710, USA.

    Figure 1.
    http://i.imgur.com/ERKv8Mi.jpg

    1. It becomes quite interesting to know what really happened to the data processing of this retracted paper after looking at the images as in the linkage above. In addition to the fig 1A to which the authors implied the possible software glitch although even the authors could not certain, it seemed like Duke was missing something in its analysis during the investigation which might lead to an arbitrary conclusion.

  8. 2017 publisher’s retraction of 2010 J Biol Chem paper by PC Kuo.

    J Biol Chem. 2010 Nov 26;285(48):37787-96. doi: 10.1074/jbc.M110.161869. Epub 2010 Oct 1.
    Osteopontin and protein kinase C regulate PDLIM2 activation and STAT1 ubiquitination in LPS-treated murine macrophages.
    Guo H1, Mi Z, Bowles DE, Bhattacharya SD, Kuo PC.
    Author information
    1Department of Surgery, Duke University Medical Center, Durham, North Carolina 27710, USA.

    Pubpeer comments: https://pubpeer.com/publications/20889505

    2017 Retraction Notice.
    http://www.jbc.org/content/292/3/1142

    “Osteopontin and protein kinase C regulate PDLIM2 activation and STAT1 ubiquitination in LPS-treated murine macrophages.
    Hongtao Guo, Zhiyong Mi, Dawn E. Bowles, Syamal D. Bhattacharya and Paul C. Kuo
    VOLUME 285 (2010) PAGES 37787–37796

    This article has been retracted by the publisher. The same data were reused to represent different experimental conditions. Specifically, lanes 2–4 and 6–8 of the actin immunoblot in Fig. 3A were duplicated. Additionally, the nuclear P-STAT1 immunoblot from Fig. 3A was reused as P-STAT1 in Fig. 3B.

    © 2017 by The American Society for Biochemistry and Molecular Biology, Inc.”

  9. 2017 publisher’s retraction of 2002 J Biol Chem paper by PC Kuo.

    J Biol Chem. 2002 Feb 15;277(7):5054-60. Epub 2001 Dec 6.
    Hepatocyte nuclear factor-4alpha mediates redox sensitivity of inducible nitric-oxide synthase gene transcription.
    Guo H1, Cai CQ, Kuo PC.
    Author information
    1Department of Surgery, Duke University Medical Center, Durham, North Carolina 27710, USA.

    Pubpeer comments: https://pubpeer.com/publications/11741883

    2017 publisher’s Retraction Notice.
    http://www.jbc.org/content/292/3/1143

    “Hepatocyte nuclear factor-4α mediates redox sensitivity of inducible nitric-oxide synthase gene transcription.
    Hongtao Guo, Charles Q. Cai and Paul C. Kuo
    VOLUME 277 (2002) PAGES 5054–5060

    This article has been retracted by the publisher. The same data were reused to represent different experimental conditions. Specifically, in Fig. 1, lane 3 was reused in lane 11, and lanes 9 and 10 were reused in 17 and 18. Additionally, lane 3 from Fig. 1 was reused in lane 2 from Fig. 3. Finally, lanes 2–4 from Fig. 1 were flipped horizontally and reused in lanes 5–7 from the left panel of Fig. 4.

    © 2017 by The American Society for Biochemistry and Molecular Biology, Inc.”

  10. 2017 publisher’s retraction of 2008 J Biol Chem paper by PC Kuo.

    J Biol Chem. 2008 Sep 12;283(37):25209-17. doi: 10.1074/jbc.M804062200. Epub 2008 Jul 2.
    Characterization of short range DNA looping in endotoxin-mediated transcription of the murine inducible nitric-oxide synthase (iNOS) gene.
    Guo H1, Mi Z, Kuo PC.
    Author information
    1Department of Surgery, Duke University Medical Center, Durham, NC 27710, USA.

    Pubpeer comments: https://pubpeer.com/publications/18596035

    2017 publisher’s Retraction Notice.
    http://www.jbc.org/content/292/3/1144

    “Characterization of short range DNA looping in endotoxin-mediated transcription of the murine inducible nitric-oxide synthase (iNOS) gene.
    Hongtao Guo, Zhiyong Mi and Paul C. Kuo
    VOLUME 283 (2008) PAGES 25209–25217

    This article has been retracted by the publisher. Fig. 2, B and C, and Fig. 4B were inappropriately manipulated.

    © 2017 by The American Society for Biochemistry and Molecular Biology, Inc.”

  11. Editorial in JBC same edition as the publisher’s retractions of the 3 PC Kuo papers above.

    Editorial
    What Happens When You Submit a Paper to JBC?
    J. Biol. Chem. 2017 292: 1535. doi:10.1074/jbc.E117.000001

    “If we receive a message questioning the integrity of data appearing in one of our published papers, we are obliged to analyze it. In cases where data integrity has been questioned, we may request original data files to compare and examine. In the most serious cases, we may ask for withdrawal of a published manuscript or print a retraction, pending the approval of the ASBMB Publications Committee. Although we at JBC are committed to tracking down bad data in submitted or published papers, we’d prefer to confidently endorse the data we’re publishing; toward that end, we are undertaking a major effort to educate potential authors about proper methods to process image data. For example, we have introduced a new series in ASBMB Today called Due Diligence (see https://www.asbmb.org/asbmbtoday/201701/DueDiligence/ for the most recent installment).”

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.