The Antibody Catalog Is Lying to You
A post-doc found manipulated images in over 17,000 commercial antibody listings. The reagents underpinning biomedical research may be far less trustworthy than anyone thought.
By The Breakthroughist Science Desk
It was Sholto David who was not looking for fraud. While searching for trustworthy data demonstrating a cell line's deficiency in the protein p53, he pulled up Thermo Fisher Scientific's online antibodies catalog and noticed something wrong with the image meant to prove the product worked. The western blot—a standard gel-based test showing whether a protein is present—had bands that were identical after flipping and rotation, as annotated by David, who first came across the image [2]. One fabricated image became a hundred.
Antibodies are, by any measure, the connective tissue of modern biological research. The immune system produces them to recognize specific proteins, and that specificity makes them essential laboratory tools [1]. Researchers use them to locate proteins inside cells, map which tissues in an organ produce a given molecule, and track whether a protein has been chemically modified [1]. They can be ordered for overnight delivery, which is part of what makes them so embedded in daily lab practice [1]. They are also, it turns out, notoriously fickle: members of YCharOS, an independent antibody validation initiative, estimated in 2024 that more than 50% of all antibodies failed in one or more applications [2].
Antibody vendors display experiment results in their online catalogs as a form of marketing—a way of showing scientists that a product will generate clean, clear data before they commit to buying it [1]. Thermo Fisher Scientific, probably the largest laboratory reagent supplier in the world, labels some of these images as "Advanced Verification" data, implying internal production [2].
The scale is what makes this more than a vendor-quality complaint. Richardson and David partially automated their detection tests and applied them across multiple catalogs [1]. The result: approximately 17,500 antibodies with problematic image changes, and nearly 7 percent of all images examined showing signs of manipulation [1]. In the Thermo Fisher catalog alone, the pair documented more than 450 images bearing signs of manipulation by June 3, 2026 [2]. A single vial of antibody solution from Thermo Fisher typically costs between 400 and 500 USD [2].
What makes this kind of misconduct so corrosive is not just the money lost. It is the way it redirects blame. When a reagent fails and the data it produces looks nothing like the catalog image, researchers do not assume the catalog lied—they assume they did something wrong.
Richardson described this dynamic plainly. "If you buy an antibody believing it will generate clean, clear data and find it doesn't, many researchers will assume the problem is them," he wrote [1]. That assumption sends scientists down a long road of troubleshooting: adjusting concentrations, changing buffers, repeating runs. "If the antibody can't produce those results, all the troubleshooting will have been a waste of time and money" [1]. Non-specific antibodies—those that bind to proteins other than the intended target—are also a major source of irreproducibility in the biomedical literature, meaning flawed reagents do not just cost individual labs time; they corrupt the published record [2].
Thermo Fisher's official response, posted in June 2026, did not concede manipulation. "The Company fully stands by the data and underlying science," it read. "In the process of preparing antibody images for publication on its website, some images may have been adjusted to clarify for presentation purposes – not to alter or misrepresent the underlying experimental results" [2]. Richardson's invitation to readers was dry: "I encourage readers to peruse the images collected in our Zenodo repository and decide what could and could not charitably be described as 'optimization for presentation and clarity'" [2].
Thermo Fisher later quietly edited its original 15-point response, removing some of the more contentious language [2]. The edits were noticed.
The manipulations Richardson found are not technically difficult to identify. They are the same kinds of image edits—duplicated bands, painted-over regions, copy-pasted background noise—that would trigger a retraction if they appeared in a peer-reviewed journal. The standards applied to published science simply were not being applied to commercial marketing. Nobody was checking.
That gap is significant. The manipulations in academic literature—the sort that get papers retracted—are subject to post-publication scrutiny from journals, watchdog blogs, and institutional investigations [1]. Catalog images exist in a different ecosystem: they are marketing copy, regulated differently, reviewed by no peer, and trusted implicitly because the alternative—validating every reagent independently before purchase—is prohibitively expensive and time-consuming. Richardson's automated pipeline changes what is possible. The same image-analysis logic that he and David applied across Thermo Fisher's catalog can, in principle, be run across any vendor's online listings [2].
Richardson published his raw repository on Zenodo and opened a Google form for other researchers to flag problematic images they encounter [2]. The crowdsourcing approach mirrors other recent scientific integrity efforts: distributed, low-cost, and difficult for any single company to suppress. Shortly after Sholto David found the first suspicious image, Johan Duchêne independently noticed a similarly suspicious image of another anti-p53 antibody in Thermo Fisher's catalog [2].
There is something clarifying about the scale of what was found. A problem this widespread does not emerge from a few bad actors making isolated decisions. It points toward structural pressures—pressures to fill catalogs, to demonstrate product performance at volume, to move vials. Whether those pressures produced intentional fraud or careless outsourcing is, in some ways, beside the point. The outcome for the researcher who spent three weeks troubleshooting an experiment that was never going to work is the same either way.
The deeper problem is that commercially-available antibodies often fail to meet the criteria of selectivity and specificity even without manipulated marketing—binding weakly to the target protein, or binding to proteins other than the intended target [2]. Faked validation data does not create that failure rate; it hides it. Scientists purchasing antibodies have no reliable way of knowing whether a product will work until they have actually bought it [2]. The verification images in vendor catalogs were meant to be the check on that uncertainty. They have not been.
Richardson's final instruction on his blog is a one-liner: "Always validate your antibodies" [2]. It reads like a standard methodological reminder. In context, it reads like a warning that the system designed to spare you that work has been telling you what you wanted to hear.
antibody images may have been optimized for presentation and clarity on the websiteThermo Fisher Scientific