You call that science? These eyes are trained to spot fakes
Bay Area image-analysis expert has a passion for research integrity
“I see scientific research as a brick wall. If one of those bricks turns out to be fake, the whole wall is unstable. It ruins the work of honest scientists who may build their research on this fake data.” — Elisabeth Bik, microbiologist and former researcher at the Stanford School of Medicine
To the untrained eye, the photo of a sliced rodent brain looks like a piece of Melba toast, with crusty scattered crumbs that seem utterly random.
But Elisabeth Bik sees something else: tiny patterns in the crust that are oddly repetitive — created, it appears, by Photoshop.
Bik’s discovery — while reading research papers on a Friday night at the dining table of her South Bay home — further heightened suspicions about the integrity of Illinois surgeon Dr. Sapan Desai, who after publishing the “Melba toast” research paper about rodent ears in 2005 went on to co-author a troubling study about hydroxychloroquine, the malaria drug hyped by President Donald Trump as a treatment for COVID-19. Two of the world’s top journals, Lancet and the New England Journal of Medicine, have since retracted the work.
For Bik, an image-analysis expert with a passion for research integrity, the case is just one of an estimated 3,000 problematic images that she’s detected in scientific papers, leading to dozens of retractions and hundreds of corrections.
With a fierce visual memory and a thirst for justice, she’s the digital Sherlock Holmes of scientific-image manipulation. From her sunny table, overlooking a garden, she devotes 12 hours a day — unpaid — to studying photos, scatter plots and other images for any evidence of wrongdoing.
Then she posts her findings online for all to see.
“I see scientific research as a brick wall ,” said Bik, a microbiologist and former researcher at the Stanford School of Medicine. “If one of those bricks turns out to be fake, the whole wall is unstable. It ruins the work of honest scientists who may build their research on this fake data.”
A self-professed introvert, she is blunt, bold and sometimes snarky in her critiques on PubPeer and her own website ScienceIntegrityDigest.com. Unlike other image checkers, who hide behind pseudonyms or work privately with journals and institutions, she usesherrealname.
Every day, she banters with her 72,000 Twitter followers, often challenging them to find fault with a posted image. Honest mistakes can happen, she notes. Often sloppiness, not fraud, is the problem.
But her work also has stirred the ire of accused researchers. To stay safe, she does not disclose the town where she lives. And her process — publicizing a discovery before privately alerting the journal or research institution — frustrates some experts.
While they admire her skill, under federal law, allegations must be thoroughly investigated, they say. Academic reputations are at stake.
If a scientist is alerted that fraud has been detected, “they may destroy the data, and that prohibits us from going back to look at whether it was valid or not,” said Lauran Qualkenbush of the U.S. Association of Research Integrity Officers.
Bik, who earned her Ph.D. in the Netherlands, has published 40 of her own papers. She has worked in the prestigious lab of Stanford’s Dr. David Relman and two biotech companies.
But in 2013, two discoveries changed her life. She was incensed to find that some of her research had been plagiarized and posted online. Later, while reading a doctoral thesis, she noticed a distinctive smudge in an image — then saw it again, flipped in a subsequent chapter, purporting to be a separate piece.
Image sleuthing became her passion. While she’s terrible at remembering human faces, she has an uncanny knack for finding patterns — whether in cell cultures, tissue samples, linear correlation plots or other images.
“In bathrooms in hotels or restaurants, I compare tile A, B and C, noticing when they’re repeated here, or rotated there,” she said. When installing a new tile floor at her home, “I drove the crew pretty mad. I knew it would make me nuts if I had to look at two identical tiles next to each other.”
Every day, working with tips she receives online, she scrutinizes an average of 100 research papers on her curved 34-inch LG monitor. She creates screenshots of each paper’s images. Then she organizes them into groups and stares at them to commit them to memory. Software can’t always detect duplicates if they’ve been cropped, rotated, flipped or altered. But Bik can.
While there’s always been misconduct in science, and image manipulation is not new, the “publish or perish” ethos is boosting pressure to produce results, she believes. This is especially true in China, where physicians’ salaries may be linked to their research. This month, she spotted identical photos in 121 different papers from China, suggesting that they were all produced by a “paper mill.”
“Once she spots something, it’s possible for another person to analyze it and see if it’s correct,” said microbiologist Dr. Ferric Fang of the University of Washington, who worked with Bik on an analysis of 20,621 papers that detected 782 duplications. “She’s a perfectionist.”
Sometimes, science selfcorrects. Since the retraction of his COVID-related research, Desai’s ResearchGate page has vanished. So has the website for his company Surgisphere.
But his “Melba toast” images endure. The University of Illinois at Chicago, where the work was done, says too much time has passed.
“It’s frustrating,” said Bik. “If you report it in a nice and polite way, it doesn’t help. Sometimes you just need to call things out.”