Ensuring that antibodies bind to a specific target and work in a particular assay is important to produce valuable research. To demonstrate the quality of their products, research reagent companies provide data on their website that demonstrates the antibody’s performance in different applications.
In the summer of 2026, data integrity researchers identified more than 450 images of antibody validation data, mostly from Thermo Fisher Scientific’s catalogue, that appeared to have been manipulated. Since then, one of those scientists, Reese Richardson at Northwestern University, has identified thousands of additional images that contain evidence of alteration. These data span 15 antibody suppliers, revealing a potential industry-wide problem.
“What we found is that every other antibody vendor on the market is selling antibodies based off of manipulated data. I think that it's really important that the scientific community knows that,” Richardson said.
An Abcam spokesperson, one of the 15 vendors implicated, said that scientists at the company conduct internal validation for their in-house generated products and review any data images provided externally. They added that when concerns arise, they conduct an investigation of the image, referring to the original data where possible and removing images where this is not available. They declined to comment on an individual example but noted that a review is underway.
A spokesperson from MilliporeSigma said that their catalogue includes internally developed and externally sourced antibodies and that they are currently conducting a review of their antibody imagery to ensure that these meet their accuracy and reliability standards. Similarly, a spokesperson from LSBio said that the company would investigate the images and source data of the western blot examples shared in an inquiry about the apparent manipulation of antibody validation data.
As of the time of publication, Thermo Fisher Scientific, ProteoGenix, and Origene—some of the other vendors mentioned in Richardson’s report—had not responded to multiple requests for comment.
Harvinder Virk, founder of “Only Good Antibodies,” a community platform that collects information about antibody characterization for researchers, found these findings “shocking” if the interpretations are accurate. He added, “I hope perhaps it's a wake-up call that we need more trusted data.”
These reports highlight ongoing concerns about antibody validation and the reliability of antibody reagents. Beyond accountability, many researchers called for dedicated efforts to address and support better antibody characterization.
Image Manipulations Are Evident, But the Intent is Less Clear
After finding the initial collection of images with apparent alterations to their backgrounds, Richardson conducted image searches with some of the photos that had similar examples of manipulation or overall formatting. This led him to multiple companies that had images with this same pattern.
Since some individuals or other manufacturers license the same product to several companies, some of these images were identical across websites, suggesting that the original antibody producer also gave the companies the validation data with the edited backgrounds. In other cases, though, when Richardson adjusted the black and white levels of the image to reveal subtle differences in the content, he found the same background used in images for different antibodies across multiple companies.
Richardson found examples of western blots of different antibodies that have similar background patterns on several vendors' websites. For example, the validation images of two products (top, anti-beta-catenin; bottom anti-HAS1) show western blots that each have a band of a different molecular weight (left). But, when the black and weight levels of the image are adjusted to reveal less intense pixels (right), a similar background pattern shows up on both blots.
Reese Richardson/ original and annotated images from the repository were grouped together by Shelby Bradford.
Richardson also identified many instances of edits that obscured portions of the image background, either with what appeared to be painting or copied and pasted sections from a different image area. “Those suggest that somebody somewhere did not like the results of the validation test and chose to fudge them, so that it looked like the antibody was performing better than intended, or better than the validation test actually showed,” he said. “In either case, somebody has been lied to, and ultimately, scientists are being lied to about these validation tests.”
Mike Rossner, an image data analyst who founded the consulting firm Image Data Integrity, agreed that the photos Richardson presented in his summary of his findings had clearly been manipulated, and that some antibody validation images share backgrounds. But he said that without the source data, it’s not possible to infer intent and cautioned against concluding that the edits imply data fabrication.
For example, he pointed out that there is a distinction between validity, which means that the antibody binds its target, versus specificity, which is when the antibody only binds that protein. In the case of western blot images with the same background, Rossner said that it’s possible that the band is real but that a company or individual felt that the image surroundings were undesirable, so they moved the real band onto a different background.
Discussing two antibodies from Thermo Fisher Scientific that shared a common background, Rossner said, “This image would certainly constitute a misrepresentation of the quality of their antibody in that case.” But he added that the current analysis alone doesn’t invalidate the antibody; it only suggests that the product may not be as specific as the image indicates.
Many antibody validation images for different antibodies from different vendors had highly similar background patterns. For example, the patterns from an anti-HCII antibody from Abcam (two left images) were very similar to that of the background from an anti-BCL11b antibody from Thermo Fisher Scientific (two right images) when the black and white levels were adjusted to reveal less intense pixels (image on the right from each company).
Reese Richardson/ original and annotated images from the repository were grouped together by Shelby Bradford.
Yet in other cases, such as multiple antibodies from ProteoGenix, the same band appears to have been copied and pasted to different locations on a common western blot background for different antibodies. Rossner said that these examples raise doubts about whether the validation analysis was performed. “That really begins to call into question the validity of the antibody,” he said, but he emphasized that it is still not possible to fully confirm that the indicated product is entirely invalid.
Rossner continued, “Looking at these things is always nuanced, and making black and white statements by just looking at a published image is fraught with problems.”
Aled Edwards, the chief executive officer at the Structural Genome Consortium (SGC), pointed out that antibody suppliers acquire many of their products, including their associated data, from other individuals. He said that currently it’s unclear whether many of the identified products are antibodies with validation data generated by the vendors or represent legacy products that entered the catalogue from another source.
“Obviously it's not good if they were modified,” Edwards said, but he added that a fairer analysis would involve classifying the products based on their origin, either internally generated or purchased from an external source. “If you want to find out the cause of these potentially egregious things, you would find out who did this manipulation, and that's a findable thing. So before [you] throw slings and arrows, find out what's going on.”
In antibody validation images from ProteoGenix, Richardson said that what appears to be the same band looks as if it has been pasted at different molecular weight locations on seemingly the same background. Here, three antibodies (left, middle, right) are shown with the western blot image from the website on the left and the same image with the black and white levels adjusted to reveal less intense pixels on the right.
Reese Richardson/ original and annotated images from the repository were grouped together by Shelby Bradford.
Richard Kahn, a cell biologist who recently retired from Emory University, began generating and characterizing antibodies before closing his lab, and he continues to be involved in the antibody validation community. Through these efforts, he has met with many industry representatives, and he found that most are interested in providing quality products.
“The optimist in me still thinks a lot of the edits to the data manipulation are mostly cleaning up to make a pretty picture.” But, he added, “They should be responsible for what they're selling and what's on their website.”
While Virk would like to know where the controversial images originated, ultimately, he does not want the focus of the issue to be on blame. “Absolutely there needs to be some accountability, and more important than the accountability is a robust plan for making things better in the future,” he said.
The Path Forward for Antibody Validation Relies Upon Research Community Support
In light of the examples of image manipulation in antibody validation data across vendors, Rossner said that his next step would be to ask antibody vendors for the source data. “It behooves these companies to present the source data underlying these images to show whether indeed the antibodies are still valid and just not as specific as shown in these images, or if in fact the antibodies are really not valid anymore,” he said.
Companies often license antibodies from other manufacturers or individuals and resell them. Richardson found identical images for the same antibody from products that may have been sold to other companies on different vendors’ sites that also contained evidence of background alteration. In this example, the western blot validation image for an antibody on Origene’s website (left two blots) are similar to the data on Thermo Fisher Scientific’s website (right two blots). The image on the left in each panel is the image as seen on each vendor’s website, while the right shows the same image after the black and white levels were adjusted to show lower intensity pixels. This analysis also showed apparent image manipulation.
Reese Richardson/ original and annotated images from the repository were grouped together by Shelby Bradford.
Rossner added that researchers have had concerns about antibody quality for decades. “I don't think this really changes that perspective. It only enhances the awareness even more,” he said. He added that the current findings further highlight how important it is for scientists to confirm the quality and application of any new reagent in their lab.
Edwards echoed this point, but noted, “Professors also have to be educated that just because an antibody is referenced in the last Nature paper doesn't mean it's good.” However, Kahn said that researchers aren’t always taught how to use and evaluate antibodies correctly and said that more emphasis should be placed on this training. Virk’s Only Good Antibodies community provides resources, including tutorials and workshops, to researchers about how to choose and evaluate antibodies to help scientists do their own internal validation.
In the short-term, Richardson agreed that researchers should be diligent about validating their antibodies in their labs and should ask their sales representatives for the original images from validation experiments to decide if they trust their quality. Considering how to address antibody validation in the long-term he argued that for-profit antibody suppliers are unreliable resources for research reagents. “It's high time that the biomedical research community considers setting up a public alternative to this essential research service,” he said. Richardson added that such an entity would be easier to regulate and hold accountable compared to private companies.
Richardson identified several different examples of background patterns that were used across antibody validation images. For example, three antibodies from Abcam (top and bottom left) and one discontinued product from SigmaAldrich(bottom right) all use a similar background that is visible when the black and white levels are adjusted (right image in each panel).
Reese Richardson/ original and annotated images from the repository were grouped together by Shelby Bradford.
Kahn and Edwards, on the other hand, advocate for more support for third-party validation communities that can partner with the existing companies to ensure the quality of these products. One group trying to fill this need is YCharOS, a community founded by the SGC and originally hosted at McGill University. In the past couple of years, YCharOS has also partnered with research groups at the University of Leicester, which Virk leads, and the University of British Columbia.
YCharOS developed a standard operating procedure alongside industry partners to characterize antibodies for specific applications.1 The protocol uses cell models in which the gene for a specific protein is knocked out using CRISPR. This removes the antibody’s intended target from the system, so researchers should see zero antibody binding. This approach is the gold standard for antibody characterization. In this partnership, companies share antibodies against specific proteins with the group, who independently test the products using the agreed upon assays.
Afterwards, Edwards said, “[Companies] can ask scientific questions, but they do not have the right to tell us not to display that in the public domain. And so, warts and all, the data are shown.” He added that 12 companies, including Abcam and Thermo Fisher Scientific, have pulled products after YCharOS demonstrated that an antibody was not reliable.2
“There're a lot of good antibodies out there. These companies have very professional pipelines to make good products that they can distribute with their commercial reach. We just need to create a nexus of characterization that can tell which of these many products that they make actually is a great performer, and that's what the public could support,” Edwards said.
Currently, YCharOS mostly receives funding to validate antibodies from disease-specific organizations, representing only a small fraction of the products on the market. Edwards and Kahn said that more public support could expand these characterization efforts further. “Eventually this will get cleaned up,” Edwards said. “Whether it get cleaned up in 20 years or five years or three is really what we have to all ask ourselves.”
Update: After publication, a spokesperson from Origene shared a press release detailing the company's findings from an internal review of an initial investigation of four images of concern, showing that the published data and original experiments agreed with each other. They acknowledged that cosmetic changes were made to remove image artifacts. The company said that, in response to concerns about image integrity, they will no longer permit cosmetic changes to antibody validation images.
- Ayoubi R, et al. A consensus platform for antibody characterization. Nat Protoc. 2025;20(6):1509-1545.
- Ayoubi R, et al. Scaling of an antibody validation procedure enables quantification of antibody performance in major research applications. eLife. 2023;12:RP91645.
