Several papers co-authored by Nobel laureate and neuroscientist Thomas Südhof are under review due to concerns raised by online commentators regarding data duplication, image discrepancies, and other issues.
Thomas Südhof Image Source: JORGE GUERRERO/AFP VIA GETTY IMAGES
Over the past year, some journals have corrected 5 papers flagged on the peer review platform PubPeer and retracted another paper. According to Science, these journals are preparing to correct at least 3 more papers. An additional 8 papers are under investigation, with authors requesting corrections or releasing new data for another 7 papers.
Currently at Stanford University in the United States, Südhof denies any wrongdoing, attributing the errors to honest mistakes and only acknowledging one "clear allegation of misconduct." He mentioned that his team's work has faced "ruthless scrutiny from individuals dedicated to exposing errors of prominent scientists on social media and blogs."
Neuroscientist Matthew Schrag from Vanderbilt University Medical Center has investigated several scientific misconduct cases and noted Südhof's unusually open approach in discussing these errors, leaving a strong impression on him. Schrag added that among the hundreds of papers Südhof co-authored, many containing extensive data, errors were somewhat expected. However, integrity expert Elisabeth Bik still expressed concerns about some papers.
Since mid-2022, PubPeer commentators, including Bik, have flagged over 30 papers by Südhof discussing how neurons communicate through synapses. These papers are unrelated to Südhof's Nobel Prize in Physiology or Medicine, which he received in 2013 for his research on how brain cells transport molecules in vesicular structures.
In response, Südhof and his team members have released raw data, high-resolution images like blots, and explanations of their work. He has also dedicated a webpage on his website to address these concerns in detail, outlining the integrity procedures followed in his lab. To preempt reviews, Südhof's co-authors recently posted 3 entries on PubPeer, pointing out errors they found in their research.
Südhof stated that most accusations are unfounded, but commentators identified 15 papers with errors. Apart from one mistake, the rest were "copy-paste" errors where spreadsheet data or images were inadvertently pasted in the wrong places, leading to their duplication or mixing images from different experiments. These errors were made by over 20 postdocs and students, with only one error affecting a paper's conclusions.
In the past year, journals including Nature Communications, Neuron, and eLife have corrected 5 papers. The Journal of Neuroscience, Proceedings of the National Academy of Sciences of the United States of America, Nature, and Science Advances are currently investigating another 8 papers. Authors have acknowledged errors in 2 of the 8 papers, including confusion and mislabeling of blots in a 2021 Nature publication and pasting incorrect images into a composite figure in a 2012 Journal of Neuroscience paper.
In March of this year, the Proceedings of the National Academy of Sciences retracted a 2023 study on the role of a protein called neurexin-2 in synaptic formation because researchers could not explain the discrepancies between original and published data. Dutch experimental physicist Maarten van Kampen was among the first to flag this paper, stating that the changes raised suspicions, suggesting, "It is almost certain in my view that the PNAS paper is based on fabricated data."
Südhof mentioned that some analyses in the papers were "improperly conducted," but there is still "a wealth of data supporting these conclusions." Neuroscientist Jacqueline Burré from Weill Cornell Medicine, a co-author on two papers under review, noted Südhof's lab for its rigorous standards, stating that "no new finding is based on a single experiment but on multiple independent experiments using different techniques, often repeated independently by multiple researchers." Nevertheless, given the extensive copying and pasting required in writing papers, "errors do happen."
In comments on PubPeer and his lab's website, Südhof pointed out that many copy-paste errors are invisible to the naked eye and can only be detected by artificial intelligence software designed to flag duplicated images. He mentioned that most erroneously copied images were meant to depict nearly identical control conditions, and accidental duplications would not impact the paper's conclusions.
Schrag emphasized Südhof's significant impact on the understanding of synapses, stating that "our field needs to address integrity-related real mistakes candidly, but it would be a grave error to overly scrutinize every mistake or disagreement as a moral issue and drive excellent scientists out of the field."
