Retraction Watch

Tracking retractions as a window into the scientific process

Authors pull Nature paper about DEET and flies

with one comment

Nature CoverAuthors have retracted a Nature paper which identified neurons that render flies sensitive to a potent insect repellent, after losing confidence in the findings. The first author, however, said she does not agree with the retraction, noting that she continues to believe the data are correct.

According to the notice, the remaining authors say they no longer support the claim that certain neurons in the antennae of fruit flies are repelled by DEET, the active ingredient in many insect repellents. The last author told us some of the paper’s results are not in doubt; nevertheless, he added, the paper would not have been published in Nature without the key conclusion, so he and most of his co-authors have pulled the paper in its entirety.

Alongside the retraction, the journal has also published a Brief Communications Arising article by scientists who were unable to reproduce the paper’s findings.

Here’s the retraction notice, published today:

We are retracting this Article because we no longer have confidence in data that support one of our key conclusions. In this Article we reported four advances in insect repellency: identification of olfactory neurons in Drosophila melanogaster that participate in repellency to N,N-diethyl-meta-toluamide (DEET); identification of an ionotropic receptor, Ir40a, expressed in these neurons required for avoidance to DEET; development of a chemical informatics method of identifying shared structural features from known behavioural repellents; and validation of a series of computationally identified natural chemicals as repellents for flies and mosquitoes. We no longer have confidence in data supporting that Ir40a is a DEET receptor. Upon reanalysis, the original calcium imaging (GCAMP) data show movement artefacts and background effects that we originally missed, which seriously undermine our confidence in Ir40a responses to DEET. In addition, Supplementary Fig. 5b presents several inappropriately re-used panels. Upon learning that A. F. Silbering et al.1 did not find defects in DEET aversion in Ir40a mutant flies, we repeated many of the original behaviour experiments. Although we confirmed significant behavioural differences in Ir40a cell-silenced flies (Ir40a-Gal4;UAS-TNTG), as reported in Fig. 2d, we have been unable to replicate observations of behavioural experiments using Ir40a-Gal4;UAS-RNAi flies. Therefore, with the exception of author Pinky Kain, we no longer have confidence in the conclusions of Figs 2, 3 and 5c, and Supplementary Fig. 5. We remain confident of the chemical informatics analyses and the identification of new repellents, which have been successfully repeated in our laboratory and by others, as reported in Figs 4, 5d and e, 6, and Supplementary Figs 2 and 6–9. Although it may still be possible that Ir40a does respond to DEET, given the issues listed above, all authors except Pinky Kain wish to retract this Article in its entirety. We deeply regret these circumstances and apologize to the scientific community.

The 2013 Nature paper, “Odour receptors and neurons for DEET and new insect repellents,” has so far been cited 42 times, according to Thomson Reuters Web of Science.

The paper — which also pointed to the possibility of new insecticides — earned some press coverage when it was released, such as from NPR and International Business Times.

Last author Anandasankar Ray, from the University of California, Riverside, told us:

The part of the paper that led to the retraction has to do with the identification of Ir40a as a DEET receptor. I was initially contacted by a colleague, who described that his lab was also characterizing Ir40a neurons and had not observed DEET responses in these neurons.

He added:

…my research group was following up with additional lines of enquiry and accumulating evidence that appeared to support a role for the Ir40a neurons. At a later time, however, it was brought to my attention that a supplementary figure (5b) had several panels that had been re-used a number of times. At that point, it became imperative that we evaluate and reanalyze all the original data and examine it carefully for sources of errors and artifacts. Upon further detailed reanalysis of the calcium imaging I found weaknesses related to movement artifacts and background effects that were not previously considered.

However, said Ray, the researcher who conducted these experiments — first author Pinky Kain — remains convinced by the data. He, therefore, asked an “experienced colleague” at another university and his other lab members to separately evaluate the data. Both lines of inquiry failed to support the original claims, added Ray. But, he noted, there is no evidence of intentional misconduct from Kain:

There was no evidence to support a finding of intentional misconduct by an institutional pre-investigation. Dr. Kain is no longer in my research group, but what she conveyed to me is that she disagrees with the retraction because she continues to stand by all of the original conclusions derived from her data.

We reached out to Kain, who is currently not affiliated with any institution. She told us:

…if you read the last 4th and 3rd line of the retraction PDF, the corresponding author said “Although it may still be possible that Ir40a does respond to DEET”. This clearly shows that even he is not confident and confuse[d] at this stage.

As part of the study, the authors created a chemical informatics algorithm that enabled them to screen more than 400,000 compounds, pinpointing some that could serve as new insecticides. As they wrote in the paper:

The candidates contain chemicals that do not dissolve plastic, are affordable and smell mildly like grapes, with three considered safe in human foods. Our findings pave the way to discover new generations of repellents that will help fight deadly insect-borne diseases worldwide.

According to Ray, this part of the paper is “not in question:”

…I want to emphasize that the chemical informatics and new repellent discovery lead by Dr. Sean Boyle are not in question, and we and others have successfully reproduced these studies several times. In fact, numerous additional new natural repellents have also emerged from this approach in subsequent work, many holding promise as affordable prophylactics against malaria, Zika, and Dengue transmission.

Nevertheless, said Ray, he and his co-authors decided to retract the paper in its entirety:

Although this finding [DEET receptor] represents only part of the study, it was a critical finding that affected the visibility and the impact of our study. The editor believed (and I agreed) that our paper was not likely to have been published as a Nature article without the Ir40a receptor part.

He noted:

And to add to that, the area of research on DEET detection mechanisms has been a highly competitive and sometimes contentious area of research with multiple papers supporting conflicting models of DEET detection leading to a lot of confusion. But there is no satisfactory mechanism yet. So from this perspective too the most appropriate course of action was to retract the paper in entirety.

Ray said he plans to post the confirmed data on the preprint server bioRxiv.

Alongside the retraction, Nature is publishing a Brief Communications Arising article (BCA) that contains the findings from other researchers who attempted to reproduce Kain et al’s results. That BCA concludes:

…our data fail to reproduce evidence for physiological activation of Ir40a neurons by DEET, or to demonstrate a genetic requirement for either Ir40a or the co-receptor Ir25a in behavioural avoidance of this repellent.  

Richard Benton, last author of the BCA from the University of Lausanne in Switzerland, added in an interview:

…after publication of the original Nature article in Oct 2013, when I discussed with Dr Ray during late 2013/early 2014 the issues of irreproducibility of the results in that work, he assured me that he “usually had 2 people set up each experiment”, that “everything that was subjective was coded and counted blind”, and that he would be “initiating a detailed review of all the raw data collected by various authors in the lab to make sure that they did not miss anything important.”

It’s not often we see BCAs published alongside retraction notices. A Nature spokesperson told us the format is used for “exceptionally interesting” or “important” comments or clarifications, adding:

Some submitted refutations may eventually be published as or alongside corrections or retractions by the original authors. Decisions about types of correction are made by the editors on a case-by-case basis, sometimes with peer-reviewers’ advice; the original authors are also consulted as part of the process.

Update: 6/22/16 4 p.m. eastern: Anandasankar Ray has now sent us the bioRxiv link to the findings he claims are “not in question” in the retracted paper. It’s available here.

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.

  • Richard Benton June 23, 2016 at 7:31 am

    A key element of my interview to Retraction Watch concerning this retraction was omitted in this article, and I append additional information on my perspective on this affair below:

    After publication of the original Nature article in Oct 2013, when I discussed with Dr Ray during late 2013/early 2014 the issues of irreproducibility of the results in that work, he assured me that he “usually had 2 people set up each experiment”, that “everything that was subjective was coded and counted blind”, and that he would be “initiating a detailed review of all the raw data collected by various authors in the lab to make sure that they did not miss anything important”.

    Despite these assurances, Dr Ray has never admitted to me the problems in the calcium imaging data (or CaLexA data which we found is also likely to suffer from background artefacts), before or since submission of our BCA in Dec 2014.

    Indeed, throughout our extensive email exchanges, he continued to defend the results in the paper, even after I explicitly pointed out the obvious movement and other artefacts in his lab’s imaging datasets that he provided to me early in 2014; these artefacts now seem to form part of the basis for the retraction.

    My co-authors and I have thus made substantial effort over more than two-and-a-half years in trying unsuccessfully to reproduce the findings of Kain et al. and to subsequently report these failures in the literature.

  • Post a comment

    Threaded commenting powered by interconnect/it code.