to BioTechniques free email alert service to receive content updates.
Statistical Error Leads to Genetics Retraction

10/04/2011
Casey McDonald

After a colleague pointed out statistical errors, the Australian authors have retracted their genetics paper.

Bookmark and Share After a colleague pointed out statistical errors in a genetics paper published last year, the University of Western Australia (UWA) authors have retracted their genetics paper.

Led by associate professor Joseph Tomkins, the UWA team published the retraction in the September 2, 2011, issue of Science (1). According to the retraction, the group made an incorrect assumption in their statistical analysis in a genetics paper published in Science in May 2010 (2). These statistical errors were brought to the attention of the group by University of Zurich research associate Erik Postma.

The cowpea weevil on its food source the black-eyed pea. Source: University of Nebraska Department of Entomology



In the paper, Tomkins and colleagues claimed to have found a negative correlation between the inbreeding depression and breeding values in the cowpea weevil Callosobruchus maculatus. This negative correlation seemed to support the mutation-selection-balance model that explains the persistence of genetic variation in populations.

But there was a flaw in their basic statistical assumption. “Their null hypothesis is wrong. Where they say they are expecting a positive correlation under the null hypothesis, what I’ve been able to show is that actually, under the null hypothesis, you really expect no correlation,” said Postma.

The UWA group failed to recognize the relationship between predicted offspring phenotypes and observed phenotypes that share additive genetic variation and are not independent.

After identifying the problem, Postma contacted Tomkins and then submitted a technical comment to Science, which was published alongside the retraction (3). In turn, the journal gave the UWA group the option to respond or retract their paper.

“It’s not the best email to get. Certainly at times it has felt quite challenging to say the least. But no one is hurt, and no one has done anything wrong in sense of a scandal,” said Tomkins.

Using a different methodology, Tomkins’ group has found a similar but weaker correlation that still supports the mutation-selection-balance model. However, since this new analysis will require a new round of peer review, the group retracted their paper.

In the mutation-selection-balance model, beneficial and deleterious mutations work together to maintain variation by balancing the pressure to eliminate variation from natural and sexual selection. “You’d expect if there’s one genotype that is best adapted to an environment, this genotype should spread through the population, and genetic variation should disappear,” says Postma.

But variation between individuals in natural populations remains. Based on their findings, the UWA researchers believes that mutations have a stronger role than random factors like drift and fluctuating selection pressures such as climate cycles.

“This is the way science works when you’re at the boundary of interesting and difficult ideas,” says Tomkins.

References:

1. Tomkins J.L., Penrose M.A., Greeff J., LeBas N.R., Retraction. Science 333:1220.

2. Tomkins J.L., Penrose M.A., Greeff J., LeBas N.R., 2010 Additive Genetic Breeding Values Correlate with the Load of Partially Deleterious Mutations. Science 328:892-894.

3. Postma E., Comment on “Additive Genetic Breeding Values Correlate with the Load of Partially Deleterious Mutations”. Science 333:1221-a.