Picture sleuths seen irregularities in figures from scores of papers co-authored by Ali Khademhosseini.Credit score: Natali Mis/Getty
In December 2024, Elisabeth Bik seen irregularities in just a few papers by a highly-cited bioengineer, Ali Khademhosseini. She began taking a look at extra publications on which he was a co-author, and the problems quickly piled up: some figures had been stitched collectively unusually, and pictures of cells and tissues had been duplicated, rotated, mirrored and generally reused and labelled otherwise.
Bik, a microbiologist and main research-integrity specialist based mostly in San Francisco, California, ended up flagging about 80 papers on PubPeer, a platform that permits researchers to evaluation papers after publication. A handful of different volunteer science sleuths discovered extra, bringing the whole to 90.
The articles had been revealed in 33 journals over 20 years and have been cited a mixed complete of 14,000 occasions. Though there are lots of of co-authors on the papers, the sleuthing effort centred on Khademhosseini, who’s a corresponding creator on about 60% of them.
Why retractions knowledge might be a robust software for cleansing up science
He and his co-authors sprang into motion. Responding to the issues, a few of which had been reported within the weblog For Higher Science, turned like a full-time job, says Khademhosseini, who till August was director and chief government of the Terasaki Institute for Biomedical Innovation in Los Angeles, California. “I alerted journals, I alerted collaborators. We tried to do our greatest to make the literature right.” In lots of instances, he and his co-authors supplied unique supply knowledge to journal editors, and the papers had been corrected.
Khademhosseini informed Nature that investigations into his work have been carried out and have discovered no proof of misconduct by him. The Terasaki Institute says that an “inner evaluation has not discovered that Dr. Khademhosseini engaged in analysis misconduct”.
The case raises questions on oversight in massive laboratories and about when a paper must be retracted and when a correction is ample. In some instances, journals have issued corrections for papers containing points that research-integrity sleuths describe as “clearly knowledge manipulation”, and the corrections had been issued with out supply knowledge. Bik and others argue that this method units a foul precedent. “I don’t assume that any a part of a examine that bears these indicators of information manipulation needs to be trusted,” says Reese Richardson, who research knowledge integrity at Northwestern College in Evanston, Illinois. He argues such papers needs to be retracted.
Khademhosseini defends the corrections and says that the conclusions of the papers nonetheless maintain. He says he has not seen any “conclusive proof” of misconduct or “purposeful manipulation” within the papers, and nothing that will require a retraction.
Discovering flaws
For 3 many years, Khademhosseini has developed biomedical applied sciences similar to organs on chips and hydrogel wound therapies. His work has been funded by the US Nationwide Institutes of Well being, and by different private and non-private companies. As a PhD pupil, he labored below Robert Langer, a famend bioengineer on the Massachusetts Institute of Expertise, in Cambridge. Khademhosseini has revealed greater than 1,000 papers, which have been cited greater than 100,000 occasions in complete. He has additionally obtained quite a few awards and honours, most lately, the 2024 Biomaterials World Affect Award, from the journal Biomaterials.
This science sleuth revealed a retraction disaster at Indian universities
In August, Khademhosseini left the Terasaki Institute. He says that he did so to work on a start-up “to speed up scientific discovery utilizing AI”, and that his departure was not associated to the PubPeer feedback. The institute additionally says that his departure “was unrelated to any analysis integrity inquiries”.
To get a way of the severity of the image-related points, Nature’s information workforce analysed the 90 papers flagged on PubPeer in session with 4 image-integrity specialists, together with Bik and Richardson, and three bioengineers. The workforce relied on steerage from the Worldwide Affiliation of Scientific, Technical and Medical Publishers (STM). This ranks picture irregularities on the idea of the kind of aberration, perceived intent and influence on analysis conclusions. Six papers had points unrelated to pictures, similar to authorship disputes or complaints about reported conflicts of curiosity. Of the remaining 84, 41 had been classed as ‘degree I’: containing minor points, similar to an by chance duplicated picture that would simply be corrected with uncooked knowledge. One other 20 papers had extra substantial issues, falling into degree II. And 23 papers thought-about degree III had extra severe points, together with a number of manipulations that immediately have an effect on the interpretation of the info. STM tips advocate that level-II and level-III points are investigated to rule out misconduct.
Khademhosseini disputes the Nature information workforce’s evaluation, saying it incorporates “substantial inaccuracies”. He says, for instance, that lots of the figures categorized as having extreme issues had been in actual fact peripheral to the conclusions of the papers, and that some errors that the evaluation deemed intentional may in actual fact have been accidents.
1000’s of extremely cited scientists have at the very least one retraction
Khademhosseini provides that the excessive variety of flagged papers outcomes from the truth that he’s revealed so many papers throughout his profession, every of which has about 80 photos. He objects to being singled out, and estimates that the speed of errors in his flagged papers is beneath the common error fee in biomedical literature, which is usually given as 4%. “In case you go after anyone who has been round a very long time, you’ll be able to dig up stuff,” he says. What’s extra, he argues that the artificial-intelligence-powered instruments that sleuths generally use to assist them detect picture irregularities elevate the typical error fee within the revealed literature, making his numbers look much more affordable.
Bik estimates that she checked out about 530 papers to seek out the errors she did. She says that she used AI to analyse a few of Khademhosseini’s papers, however that she additionally verifies every irregularity by eye.