The authors of two studies on the gut microbiome’s influence in Parkinson’s disease and anxiety published in Cell and Nature, respectively, are working to correct data duplications flagged by a software engineer and commenters on PubPeer.
“The issues are inadvertent errors, and we are already working with the journal[s] to correct them,” Sarkis Mazmanian, professor of microbiology at the California Institute of Technology and an investigator on both studies, wrote in an email to The Transmitter. “Importantly, they do not alter any outcomes, interpretations, or conclusions in the published articles.”
Dorothy Bishop, emeritus professor of developmental neuropsychology at the University of Oxford, disagrees. The duplications cast doubt on the data, reduce confidence in the studies and add to “the general impression that this literature on the microbiome and animal models is not of a very high standard,” says Bishop, who has written about issues in research on autism and the gut microbiome.
The Parkinson’s disease study, published in Cell in 2016, used mouse behavioral data of motor function that has identical sets of numbers in two different experimental groups. Markus Englund, a software engineer, found the overlap when using his software to identify duplications in Dryad, an open-source research data repository. The 2016 dataset contains so many identical numbers in a row “that you wouldn’t expect to ever see this by chance,” Englund says.
Ben Woden, a collaborator of Englund’s, posted the concerns on PubPeer in January, and Englund reached out to Dryad and Cell to inform them of his findings, Englund says. Dryad posted an expression of concern on the dataset, and Cell is “investigating the matter,” wrote Cell editor-in-chief John W. Pham in an email to The Transmitter.
Similarly, the anxiety study in question, published in 2022 in Nature, which examined how microbes affect anxiety-like behaviors in mice, has identical data points in a spreadsheet, according to anonymous comments on PubPeer. Englund says he agrees with the finding: “It’s another case of duplicated mouse behavioral data.” Nature is “looking into this matter carefully following an established process,” says Francesca Cesari, chief biological, clinical and social sciences editor at the journal.
It is not the first time Mazmanian’s work has come under scrutiny. A 2019 study showing that microbes from people with autism could alter behavior in mice was criticized for its statistical analysis, The Transmitter (then Spectrum) reported. Mazmanian is co-founder of, and board member and scientific adviser for, Vertero (previously Axial Therapeutics), a Massachusetts-based company that tests microbiome-derived treatments for neurodegenerative and other diseases.
T
In this study, mice that overexpressed alpha-synuclein took significantly longer to remove a sticker from their nasal bridge than did wildtype mice, a measurement of motor function called adhesive removal time. Treating these mice with antibiotics to get rid of the microbiome dampened the detrimental effect of alpha-synuclein on motor function, whereas recolonizing the gut of previously germ-free mice exacerbated it.
But half of the 10 data points for the recolonized wildtype and alpha-synuclein groups are identical, Englund found. And some wildtype mice removed the tape in about 0.5 seconds, which does not seem feasible, says Mu Yang, data sleuth and director of the Mouse NeuroBehavior Core at Columbia University. (Some mice remove the adhesive immediately after being put in the cage, resulting in these short times, explained study investigator Timothy Sampson, associate professor of cell biology at Emory University, in an email to The Transmitter.) And other posters on PubPeer raised concerns about the statistical methods, including the small sample size.
The Nature study, which has been cited more than 400 times and was covered by The Transmitter, found that a microbial metabolite called metabolite 4-ethylphenyl sulfate impaired myelination and led to anxiety-like behaviors in mice. One of the main figures in the paper, which shows that mice treated with 4-ethylphenyl sulfate spend less time in the more exposed area in an open field test than wildtype mice, contained duplicated numbers in both groups, according to a comment on PubPeer.
“In some cases, re-analyzing the data with duplicates removed still gives a significant comparison,” the anonymous poster wrote. However, they added, “the p-value is not as low.” In an email to The Transmitter, study investigator Brittany Needham, assistant professor of anatomy, cell biology and physiology at Indiana University School of Medicine, confirmed that the conclusions of the data were not altered when she and her colleagues replaced the duplicates with the correct values.
