Antibodies form the backbone of many scientific pursuits. From western blots to immunohistochemistry to clinical diagnostics, these Y-shaped proteins are vital to research groups across disciplines.
In addition to their star role in many experiments, these products are also expensive. This makes choosing the correct antibody to get the job done doubly important. To do this, scientists rely on validation data provided by companies to make sure they buy the right product and have a reference of what results to expect.
When analytical scientist Sholto David at OXB, a pharmaceutical company, found a suspicious western blot in Thermo Fisher Scientific’s validation data for one antibody, he shared the image with colleagues on social media, thinking it was a one-off. Then a researcher commented that the data for another antibody had a similar problem.
“I just thought, ‘Where there's smoke, there's fire.’ And I spent an hour looking, and I found 10 more,” said scientific disinformation researcher Reese Richardson at Northwestern University. Intrigued, he studied images with similar file names to look for more examples of image manipulation. “That turned out to be a rich vein because I just kept going and just kept finding things,” he said.
Richardson shared these findings in a blog post and on social media. The response from researchers came in swiftly, expressing outrage, disappointment, and general concern. In an animated gif, multiple images of different antibodies have areas with the same background pattern. Jim Woodgett, a cell and molecular biologist at the Lunenfeld-Tanenbaum Research Institute called this “stunning.”
Richardson and David started a repository to make researchers aware of this issue and for others to share images of concern that the two are continuing to review. As of June 4, 2026, this catalogue hosted 462 images from validation data, many from Thermo Fisher Scientific, that contain evidence of manipulation. Richardson said more companies could be added as he and David process submissions.
Image Manipulation Evidence Breaks Researchers Trust with Distributors
The alterations that Richardson and David found included examples of copying and repositioning bands, cropping out and pasting in different backgrounds, and even using tools to paint over or obscure regions of the blot. “As soon as I saw [painting on the blot], I had to sit back and just try to stop myself from losing my mind. There is no circumstance under which this would happen in an image naturally from compression [or] artifacts from bubbles in the western blot,” Richardson said, adding that the findings were “maddening.”
Richardson identified examples where background elements appeared to have been obscured or painted out on some validation data images (right), since the features would not occur naturally in a western blot. The image available online (left) shows no evidence of this alteration.
Richardson, R., & David, S. (2026)
Indeed, Richardson’s posts on social media garnered multiple responses, most expressing shock and disappointment. “I was not surprised that the antibodies are bad,” said Paul Jenkins, a neuroscientist at the University of Michigan. “What I was surprised by is that those blots appear to be fabricated.”
The biggest cause for alarm amongst researchers was that these manipulations are in data that researchers use to determine the products’ specificity and to select which product to buy, which directly ties into researcher time and lab costs. “Knowing that those images might not correspond to the result that the company obtained causes significant concern to me,” said Marcelo Bonini, a cancer biologist at Moffitt Cancer Center.
A representative from Abcam, another antibody supplier who had a product identified with altered validation images, shared in an emailed response from the company, “We are reviewing the concerns raised through our scientific and quality processes and will take appropriate action where needed. We remain committed to high standards, transparency, and providing clear, reliable data that researchers can trust.”
Thermo Fisher Scientific had not responded to a request for comment as of the time of publication. On their website, Thermo Fisher Scientific currently has a page with frequently asked questions (FAQ) about these claims. Under one of these, they indicate that some images in their catalogue “may have been edited, optimized or otherwise adjusted.” On their product pages, a note with this information indicates that these changes were made for presentation purposes.
“I am completely unsatisfied with their answer that these images were manipulated for online display purposes. That doesn't make sense to me whatsoever,” Jenkins said, adding that he felt that the company was trying to gloss over the issue.
Bryan Jones, a retinal neuroscientist at the University of Pittsburgh, said, “There’s clear image manipulation, so telling us that the king has no clothes, it’s just not working.” He added, “To have a product so clearly being falsely represented is a problem.”
Woodgett also found Thermo Fisher Scientific’s response disappointing, saying that it disregarded the scientific community’s concerns and implied to young researchers that altering data in this way is appropriate. “It’s not tolerable at all. It's not, as it just sends exactly the wrong message,” he said. “There's no excuse for it, to tidy [the blot] up. If you think it's embarrassing, just do the blot again.”
Jones added that, when scientists do edit images for presentation purposes, such as by adjusting the contrast or pixel intensity, the alteration is applied across the entire image, unlike what is displayed in the repository examples. They also clearly disclose these changes and have the original image available upon request. He and other researchers said that, if reagent suppliers are editing their validation data for online viewing, they should also be disclosing that information.
Unreliable Antibodies Have Time and Cost Consequences for Researchers
While this is the first time that there have been issues found with a company’s validation data involving image manipulation, at least that became public, problems with antibodies have plagued researchers for years. In 2023, one study that analyzed 614 antibodies showed that, against common specificity tests, 50 percent failed at least one assessment.1 While this study did not make claims about unreliable vendor data, it emphasized the challenges researchers experience with these products.
The background in one antibody from Abcam (HC ii) shares a pattern with several other antibodies identified in the repository (right). When the contrast is raised from the original image (left) shadows and black marks appear.
Richardson, R., & David, S. (2026);original and annotated images from the repository were grouped together and image borders added by Shelby Bradford.
Jenkins said that part of the reason why antibodies have so many issues is because they can be hard to produce and that companies often don’t have the resources to verify them against the gold standard, a knockout model. With this resource, scientists can confirm that an antibody signal disappears with the absence of the target protein, confirming its specificity. Additionally, some suppliers do not provide the antigen that was used to create the antibody, so researchers cannot confirm for themselves if the product matches their target before they purchase it. This, Jenkins said, is why reliable validation data is even more important.
Indeed, Bonini said that the images in the catalogue not reflecting the results that the company obtained in the first place is misleading. While he acknowledged that some variability between antibody batches is normal, he said issues that researchers experience as a result of inaccurate validation data could be blamed on this variability or user error erroneously, making it harder to obtain refunds. Some companies even require that researchers conduct specific experiments to prove that the antibody is in fact faulty before they will issue a reimbursement, adding to lab expenses.
Even when researchers do jump through these confirmation hoops, they have still struggled to get responses from companies. Jenkins found that an antibody from Sigma-Aldrich, today owned by Millipore Sigma, was exhibiting nonspecific binding. To his frustration, even when he submitted a report to the company with evidence about the problem, he said he never heard back and that, last he checked, the antibody was still on the website. As of the time of publication, Millipore Sigma had not responded to a request for comment about this situation.
The current scenario, though, Jenkins said is more concerning. “I feel like this goes well beyond the worry about, ‘Hey, this one wasn't well-validated’ and into a place where ‘Are these blots even real?’ I mean, it's millions of dollars of taxpayer money lost.”
Several researchers said that they have invested thousands to tens of thousands of dollars in antibodies over their careers. Seeing evidence of unethical image manipulation rattled many researchers’ trust in reagent companies.
“It makes you wonder how much time, expense, and effort, and trainee time is wasted through following up poorly performed research, and there's really no excuse. So, I think we do have to hold companies who are selling reagents to a high standard, and they should be holding themselves up to the same standard,” Woodgett said.
Bonini agreed, “It's incredibly frustrating when you realize that you didn't get a piece of data or you didn't move your science forward, not because you didn't plan an experiment accurately or because there are pieces that you don't know, but because the product was faulty.”
Researchers Call on Vendors to Make Amends for Errors
Moving forward, the researchers said that reagent suppliers should make transparency a key priority. This could start with indicating that images may have been edited, preferably with details about the changes, and making unedited versions available. In addition to adding a provisional statement about image edits to their catalogue, Thermo Fisher Scientific says on their FAQ page that raw, unedited images are available upon request.
Yet, for some scientists, the damage may have already been done. “Frankly, for me, I will strongly consider whether I'll ever order antibodies from them again,” Jenkins said.
Several other researchers agreed that a company’s reputation is a large factor in the decision to purchase products from them. Thermo Fisher Scientific’s response to this incident, they said, has not inspired confidence among scientists.
Richardson and Woodgett both said that one way to begin to resolve this incident is for these companies to work with external antibody validators or researchers experienced with the protein being targeted, including the relevant models in which to test the antibodies. Indeed, Jenkins said that he and his team worked with a small company in the past to help them validate their products.
On social media, Bonini remarked that researchers who purchased antibodies with seemingly manipulated validation images should submit refund requests. While he said that the comment was in jest, he said that if he were personally affected, “I would love to have a refund.” He added that he thinks there should be some form of reparations made available.
Given the evidence presented that at least a fraction of the antibodies from Thermo Fisher Scientific, and possibly more companies, could have different specificities than what is presented on the website, Bonini said, “I'm concerned, but what am I going to do about it? Because I still depend on these tools.”
As one solution, Bonini said that his group will be adding more controls to their antibody evaluations. He said that while this will increase his team’s time and internal costs, “As a steward of the public's money I need to be accurate in my research even if companies and providers fail to meet those standards. Ultimately, I'm responsible for my research.”
- Ayoubi R, et al. Scaling of an antibody validation procedure enables quantification of antibody performance in major research applications. eLife. 2023;12:RP91645.
