From a small town in Wales, a scientific sleuth has shaken Dana-Farber — and elevated the issue of research integrity

PONTYPRIDD, Wales — The blog post that has shaken the leadership of Boston’s Dana-Farber Cancer Institute, one of the world’s preeminent cancer research centers, was written some 3,000 miles away, in a bare-walled, sparsely decorated flat, save for a stack of statistics books and a collection of Rubik’s Cubes.

It’s here that Sholto David, an unemployed scientist with a doctorate in cell and molecular biology, spends his time poring over research papers looking for images with clues that they’ve been manipulated in some way to portray misleading findings — perhaps duplicated, spliced or cropped, or partially obscured.

As he’s toiled away over the past three years, often long past midnight, he’s flagged issues on more than 2,000 papers on a site called PubPeer, where researchers can critique and discuss published studies. His comments are sometimes met by a study’s author dodging the questions raised, and sometimes result in a correction or retraction. Often though, they’re met with no response.

But earlier this month, David helped ignite a furor after publishing a blog post that outlined purported errors he and other researchers noticed with images in dozens of papers from top Dana-Farber researchers, including the institute’s chief executive, COO, and research integrity officer. His tone was mocking, at times scathing: A paper co-authored by CEO Laurie Glimcher “includes some impressive contributions to art, but perhaps not to science,” he wrote; he described papers as “calamitous” or an “epic fail.” Some of the manipulations, David alleged, seemed deliberate. “Dana-Farberications,” the headline called them.

On Monday, three weeks after the blog post was published, Harvard-affiliated Dana-Farber said it has plans to retract six of the papers and correct 31 of them.

David’s Jan. 2 post on the For Better Science blog was his latest foray into challenging what he sees as inferior research. On his YouTube channel, where he dissects problems in papers and with alternative medicine practitioners, he has episodes of “Science Police,” the opening of which shows him running down a trail and popping out of a river in a police vest and hat, like the geekiest spoof of “COPS” imaginable. With cheesy graphics, he puts authors who’ve had to retract papers for poor research in “science prison.”

The intensity with which David approaches tasks is readily apparent. At his flat, while showing a reporter how he detects altered images of mice on his dual-monitor setup, he joked — was he really joking? — that he could discuss these images for hours. On another table, he was building a model of his favorite plane, an F-4B Phantom, from a kit that a friend had given to him; he wanted it to be known that he knew he was not doing a great job assembling and painting it, as if such a thing would be clear to an average person.

David, 32, is part of a lineage of scientific sleuths dating back now more than a decade who — often outside their day jobs or in retirement — comb through papers, sniffing out signs of shoddy data analyses or image chicanery. They alert journal editors, researchers, and institutions, or post about it on PubPeer. And while some of their findings don’t get addressed, occasionally the study in question will quietly join the thousands of others — 10,000 alone last year — that are retracted annually for any number of reasons.

But sometimes, perhaps because they involve institutions like Harvard, or are linked to senior figures, or because they come on the heels of the resignation of Harvard’s president in part over plagiarism questions, allegations of research misconduct become international news. Last year, in another prominent case, Marc Tessier-Lavigne resigned as Stanford’s president after image manipulation investigations into papers he co-authored (though he wasn’t found to have manipulated data himself).

After David published his blog post, his claims were covered by the Harvard Crimson and then STAT. Since then, he’s been interviewed by outlets including Nature, the Wall Street Journal, and the New York Times.

David acknowledges that he is doing the type of digging that others do, and that he’s been doing it for less time than many other moonlighting reviewers. But David, who is more brash than many of his fellow sleuths, now finds himself in the middle of a kerfuffle with people wanting to know what drives him to do this work and his thoughts on how to solve research integrity problems, which he says he doesn’t have any good big-picture insights about. He’s also faced scrutiny over the language and tone he uses in his posts and the admittedly crude Photoshopped images he makes, including one with a photo of Irene Ghobrial, one of the Dana-Farber researchers involved, cropped onto the body of a Barbie doll.

“The way that post was written — it’s puerile, it’s snarky, it’s misogynistic,” Barrett Rollins, Dana-Farber’s research integrity officer, told STAT this week about David’s blog. “I don’t want to go too deeply into this, but that was really upsetting,” adding that “to Photoshop [Ghobrial’s] picture onto Barbie’s body is inexcusable.”

At the same time, after the link to David’s Patreon account appeared in the Times, the number of contributors has gone from one to 80.

It’s all a bit uncomfortable, David said over a veggie burger on Friday afternoon at a pub in Pontypridd, a small town set into the hills 30 minutes outside Cardiff that was once a coal mining hub. What matters, he said, is that errors get corrected, and he hopes what he’s doing contributes to some greater good.

He also seems to want to expose how scientists in lofty positions, making big salaries and attracting large grants, have gotten there even though their papers have such evident problems that he notices them in a matter of minutes. He’s alleged research shortcomings by authors who’ve held positions at the National Cancer Institute, Memorial Sloan Kettering Cancer Center, and Queen Mary University of London.

But the fact is, David does this work because he simply enjoys it, enough to stay up until 2 a.m. reading papers even when he was employed. He views it like a puzzle, like some sort of 2D version of the Rubik’s Cubes sitting in his flat — can you look at this scan of two dozen mice and spot which, if any, of the animals have been copy-and-pasted?

“I’m not a successful researcher, am I? Let’s be realistic,” he said. “But I’m seeing all these people who’ve got these high-flying careers and they’re just bloody Photoshopping all the blots. Wouldn’t you be mad about that? I didn’t go far in academia, and probably because I’m not a very inventive, creative, or prolific scientist in a lot of ways. But I’m just fascinated by it.”

David tallies the results he’s elicited, which can give the impression he’s hunting for sport, racking up retractions to fill a trophy case rather than serving as a bastion of scientific standards. He also admits to sending rude emails and writing in ways that are designed to provoke. But that’s in part, he said, because his past efforts at playing polite and going through usual channels — writing letters to journal editors or contacting researchers — didn’t get results.

His approach worked just fine when he investigated in relative obscurity. But David is anxious about the idea that, as more people learn who he is, they might think that he’s some zealous crusader for scientific purity, or that they might contribute to his Patreon as if doing so is an act of standing up for careful inquiry or supporting something he’s doing full time. He’s just as likely to spend his hours gaming, and in the past he’s cut out for weeks at a time to go on an epic bike ride, including pedaling from Amsterdam to Budapest and Florida to Texas. He’s wondering if he’s going to have to start moderating himself now.

“I’m the sort of person who writes 2,000 comments on some obscure website just for doing it,” he said. “I like a little bit of engagement, but I would do it anyway.”

He also worried that any focus on him would in a way repeat the mistake that occurs in science, when “there are famous names writing famous papers at famous institutions” — and in turn people are more likely to accept their findings or not give them the scrutiny they deserve. Why should any attention be on him, not on the errors in the papers?

“You can learn all you want about me, but it’s not going to change what’s true about the papers,” he said.

At the heart of the matter, then, David has largely been proven right — though Dana-Farber’s ongoing investigation has not yet determined who was at fault, nor whether any of the errors were deliberate. While the institute has said that three of the flagged papers don’t need corrective action, the vast majority did. Even as Dana-Farber’s Rollins criticized David’s blog this week in his conversation with STAT, he also said, “The substance of it is something that’s serious and needs to be addressed.”

David knew some people wouldn’t like the Barbie Photoshop, but he was also ready to explain.

One of the other Dana-Farber researchers under scrutiny is Kenneth Anderson, whose face David Photoshopped onto a Ken doll in the same image that featured Ghobrial as Barbie. David also noted that in a previous blog post about a different researcher at a different institution also named Ken, he had Photoshopped that Ken’s face onto Ryan Gosling-as-Ken’s body. On his YouTube videos, he has for years doctored images of the researchers whose work he’s criticizing as clowns or wearing propeller hats.

“Because I don’t have a massive imagination, and because the ‘Barbie’ movie is huge, I was like, I’ll do the Ken thing again,” he said.

David started thinking about research integrity, or really, the poor research that makes it into publications, when he was in his Ph.D. program at Newcastle University, where he focused on a type of bacterial protein. He planned to write a systematic review, and to figure out how to go about doing so, he decided to follow the methods in a published review and try to replicate it. But as he worked through the process, he started finding mistakes in the published one.

After he finished his doctoral program, David worked for a few years at Oxford Biomedica, a gene and cell therapy company. He then moved to Pontypridd for a job at a nearby contract research organization, but quit last year after a few months and finding that it wasn’t a good fit. He’s been living on his savings, and will eventually have to find a job, ideally in an industry role, he said.

David’s budding interest in finding research errors coincided with the growth of scientific sleuthing. The field started about two decades ago, as journals became digitized, and people could fish for fraud from home, instead of trekking to a medical school library.

The sleuths have over time branched into their own specialties: some hunt for plagiarism, others for flaws in statistical analyses, and others — like David — for massaged images. Those focused on image manipulation have typically caught problems by eye, but some have also started to use software like the AI-based ImageTwin.

Ivan Oransky, the co-founder of the Retraction Watch blog (and an occasional contributor to STAT), said the increase in retractions, from a few hundred a year 20 years ago to thousands a year now, is largely attributable to sleuths. Part of that is their direct work in flagging problematic papers, but part of it is indirect, too, with some publishers creating systems to better catch flawed papers after they realized people were paying attention.

“Without sleuths, I don’t think you’d see a fraction of the retractions,” Oransky said. “You don’t find these problems if you’re not looking, and no one was looking.”

The growing awareness about how common manipulated images are in papers has raised questions about how the drive to publish compelling research can lead to scientists taking shortcuts. It’s also put a harsh spotlight on how research is conducted and how much, or little, oversight lab leaders are providing over trainees. Graduate students and postdoctoral fellows do the actual brunt of the work, and may feel pressure to generate results that will appeal to their bosses.

One aspect that’s been lost in some coverage of the Dana-Farber retractions — and which David is quick to bring up — is that other sleuths had flagged some of these papers years ago on PubPeer. His blog post, in addition to detailing his recent claims, also summarized those earlier findings.

And in fact, Dana-Farber has said since the media coverage began that it already had a review underway before David started commenting on the research. That could help explain why within just a few weeks, the institute had announced it was moving to retract and correct papers. Typically, such reviews take months.

Some have credited Dana-Farber for its lack of defensiveness and its quick effort to amend errors. But from the outside, the timing seemed to raise the possibility that the blog post, and subsequent media attention, may have forced Dana-Farber’s hand.

Dana-Farber said its scientists “have been reviewing some of these papers for months now, which I suppose on one level is a good thing, but also makes you wonder just how vigorously that investigation has been moving,” the influential science blogger Derek Lowe wrote this week, noting that, “For Better Science blew the whistle here, as they have on many other such situations.”

In their comments about possible research misconduct, many sleuths write carefully. They present their claims in a just-the-facts way, avoiding words like “fraud,” which can imply intent, rather than a lack of oversight, simple mistakes, or more anodyne explanations.

Part of the idea is to keep attention on the crux of the issues being raised. Oransky said there have been cases in which researchers whose work is questioned try to discount legitimate claims by focusing on the inflammatory or offensive language with which the allegations are made.

Elisabeth Bik, perhaps the most prominent of image sleuths, said she broadly agreed with the claims David made in his blog post. But in her work, she steers clear of making things personal, she said.

“He has used a tone I would not use myself, let’s put it like that,” she said.

David said he aims to be purely academic in his posts on PubPeer, but he sees blogs as a different arena. The For Better Science site is filled with cartoons and irreverent writing. It’s meant to be entertaining for people who care about this sort of thing.

David also admits to sending rude emails to journal editors, but that’s in part because they don’t typically reply, so why not launch into them?

“I would like to generate a response, and the way you generate a response is not, unfortunately, in science by writing polite emails and by asking very nicely,” he said. “The way you might get people to respond is by calling them out, saying this is stupid, and a lot of these errors are stupid.”

The measured tone some sleuths employ isn’t just a rhetorical strategy. It’s also to avoid being sued.

Last year, after a blog called Data Colada outlined allegations of research fraud against Harvard Business School professor Francesca Gino, Gino filed a lawsuit against the sleuths behind the site, arguing they had defamed her. Such lawsuits might not succeed, but they could deter others from raising the concerns they spot in research, those in the field say.

The legal risks that sleuths take on underscore the gaps in enforcing the standards of scientific papers. Peer review by other researchers, the standard way journals vet submissions before deciding to publish, focuses more broadly on scientific rigor and originality, and is not intended to catch data discrepancies or outright fraud. Academic publishers don’t have effective systems in place to police themselves, and instead have wound up relying on the passion and labor of outsiders to clean up their messes. Yet as crucial as the sleuths’ work has become in forcing higher-quality research, they’re doing the work for free and even leaving themselves vulnerable to lawsuits.

For his part, David does not seem overly concerned about his posts resulting in him getting sued. As for the Dana-Farber case, he said he’s just pleased that the institute has moved to correct the errors that he and others pointed out.

Amid all the attention, David has decided to say yes to the journalists’ requests he’s been receiving, including being photographed. And yet that tension remains: He’s leery about any spotlight on him as an individual, and whether it will distract from his actual points.

“Images are an important part of this, but not images of me,” he said.

Angus Chen and Jonathan Wosen contributed reporting.