Sometime last year, the University of Zurich’s Erik Postma was reading a paper in Science titled “Additive Genetic Breeding Values Correlate with the Load of Partially Deleterious Mutations” when he realized something.
The authors, led by Joseph Tomkins of the University of Western Australia, had made a mistake.
Postma set to writing a “Technical Comment,” the way that Science usually deals with criticism of a paper’s methods. He sent his first draft to the Tomkins group:
I told them that I thought that I found a mistake in their analyses and that I was about to submit a technical comment. However, that I would be more than happy if they could show that I was wrong. I happen to know some of the authors quite well, and they took my email in good spirit and took my criticisms very seriously. We then emailed back and forth a bit more and I did a few more analyses that further confirmed my suspicions.
A TECHNICAL COMMENT BY POSTMA (1) HAS CORRECTLY IDENTIFIED A PROBLEM WITH OUR 14 May 2010 Report, “Additive genetic breeding values correlate with the load of partially deleterious mutations” (2). There was an error in the methodology we developed to generate the null hypothesis against which to test the prediction that inbreeding depression correlates with breeding values in cow-pea weevils. Our mistake was to overlook the genetic relationship between predicted offspring phenotypes (p) and observed phenotypes (o). Specifically, p and o are not independent but share additive genetic variation. The null hypothesis we generated did not take this into account and hence the magnitude of the reported negative correlation between breeding values and inbreeding depression was overestimated. We reanalyzed our data with a different methodology and found a similar, but weaker, overall effect for the negative relationship between inbreeding depression and breeding values. Because the published correlation values are in error, we retract our original Report.
The original paper was published in May 2010, and has been cited seven times, according to Thomson Scientific’s Web of Knowledge. The Technical Comment is available for free, but the retraction notice itself is behind a paywall.
It took more than seven months for the journal to accept the letter, and then about two months to publish it. That seemed a bit long to us, and to Postma:
Although I can only speculate about why the whole process took so long, I suspect it was at least partly because Science had to decide on whether they would let Tomkins et al reply to my comment (as is usually the case), or if they would ask them to retract their paper.
It should be noted that Science typically doesn’t accept Technical Comments that respond to papers published more than 6 months ago. When I submitted mine, however, already a bit more than six months had passed. Science could thus have easily decided not to consider it. The fact that they did shows to me that they do take these kind of things seriously.
Postma understands the allure of the original findings. In his letter, he writes (link added):
The idea of a negative correlation between additive genetic breeding values and mutational loads is appealing because it provides an answer to the enigmatic question of how genetic variation is maintained in the face of selection (13).
Postma and Tomkins both said that no other papers would need to be retracted because of the error, which Tomkins called an “unfortunate mistake.” Tomkins quoted a letter sent to him by Science editor in chief Bruce Alberts and deputy editor Andrew Sugden:
Please rest assured that in requesting retraction we are not in any sense casting doubt on your integrity as researchers, and we fully accept that this situation has arisen through honest error.
Postma also had reassuring things to say about the whole episode:
I would like to emphasise again that I can say nothing but good things about how Tomkins et al dealt with what is probably everyone’s worst nightmare. Furthermore, I think that the same thing can be said about how Science dealt with this.
Hat tip: John Girdlestone