The scale of the problem is not trivial. In one large-scale test, more than half of the antibodies examined were found to be unfit for purpose in at least one of their potential applications. Examples abound in the literature: a 2022 study showed that murine bone marrow macrophages and human monocytes do not express atypical chemokine receptor 1, contradicting earlier reports built on antibody staining, and a 2024 analysis argued that Bax detection across more than 1,400 publications might be flawed because of unreliable reagents. When antibodies fail silently, the consequences ripple outward: incorrect results, wasted time and money, unnecessary animal and human tissue use, and a measurable contribution to the reproducibility crisis that has troubled biomedical research for over a decade.

Concerns about antibody specificity were raised loudly as early as 2015, when researchers estimated that unreliable antibodies were wasting roughly 800 million dollars worldwide every year and called for standardized reagents. The International Working Group for Antibody Validation subsequently proposed five pillars of antibody validation, stressing that validation must be application-specific, since an antibody that works well in a Western blot may fail completely in immunohistochemistry. Yet despite this growing awareness, insufficiently validated antibodies remain in widespread use. Newer initiatives are trying to change that. One of the most prominent, YCharOS, short for Antibody Characterization through Open Science, systematically tests commercially available antibodies using knockout samples, the gold-standard approach for assessing specificity. A signal in a knockout sample, where the target protein has been genetically removed, indicates that the antibody is binding something other than its intended target. By making such data openly available, YCharOS aims to shift validation from an individual laboratory’s troubleshooting exercise toward a shared community resource.

Two complementary studies published in the same issue of PLOS Biology now tackle the question of why the problem persists and what can realistically be done about it. The first, by Biddle and colleagues, quantifies the damage. Drawing on data from the YCharOS platform, the researchers examined 614 antibodies characterized across multiple research applications and identified 97 that failed under every condition tested. They then traced these failed reagents into the scientific literature, analyzing 785 publications in which the antibodies had been used and identifying thousands of animal and human biological samples that had been studied without any reported evidence of validation.

The extrapolation from those findings is sobering. Biddle and colleagues estimate that millions of animal and human tissue samples may have been consumed in experiments involving antibodies that would fail independent testing. Behind each failed antibody lies a chain of experiments, grants, doctoral theses, and animal lives that produced results of uncertain meaning. The study also probed the human side of the problem, surveying scientists about how they choose antibodies. The answers revealed that selection is often skewed by previous use in the laboratory, by what has already been published, and by supplier reputation, rather than by a systematic evaluation of the reagent’s actual performance. In other words, researchers frequently rely on social and commercial signals that correlate poorly with analytical validity.

Identifying a problem is important; providing solutions matters even more. That is the contribution of the second study, by Blades and colleagues, which recognizes a structural feature of the dilemma: the problem cannot be solved simply by asking individual researchers to validate their antibodies more rigorously, and responsibility is distributed across many actors who each tend to assume that someone else will address it. To break that deadlock, the authors conducted a Delphi consensus process, a structured method for building agreement among experts through iterative rounds of evaluation, involving 32 international experts drawn from five stakeholder groups: academic researchers, scientific publishers, research funders, antibody manufacturers, and institutional research leaders.

From a starting list of 33 proposed actions, the participants identified 15 that were judged both effective and feasible to implement by 2030. Crucially, these actions distribute responsibility across the entire research ecosystem rather than piling it onto bench scientists alone. Researchers need access to appropriate training and validation resources; institutions can embed validation standards in core facilities and hiring practices; funders can require validation plans in grant applications; publishers can demand documentation of antibody performance in manuscripts; and manufacturers can supply knockout-validated data with their products. The study thereby moves beyond the generic exhortation that antibodies should be better validated and toward something closer to a roadmap, in which each stakeholder holds a distinct and defined role in improving reagent reliability.

Whether that roadmap translates into changed practice remains an open question. Consensus among stakeholders is a necessary first step, but implementation will demand sustained commitment and coordination across a research system that is notoriously fragmented and slow to alter its habits. The authors of the Primer argue that the next challenge is to make rigorous validation part of routine biomedical research rather than an occasional heroic effort. That challenge may grow more urgent as new technologies reshape how science is done. Artificial intelligence is increasingly used to search and synthesize the scientific literature, but such systems lack critical judgment and do not exclude findings based on poorly validated reagents, making their outputs potentially unreliable. Erroneous conclusions baked into the literature could then be propagated into new analyses, recommendations, and hypotheses at machine speed. Ensuring that future AI tools are trained to account for reagent quality is, in this view, an essential safeguard.

At the same time, artificial intelligence is not only a risk but also a potential source of alternatives. Computational protein design has advanced rapidly, with systems now able to generate protein binders from a target structure alone and tools such as BindCraft enabling one-shot design of functional binders. These designed molecules could complement or even replace conventional antibodies in some applications. But their success, Duchêne notes, will depend on whether the lessons learned from decades of antibody validation are taken into consideration from the start, and on systematic testing of whether the new binders specifically recognize their intended targets. A new generation of reagents that repeats the old mistakes would simply relocate the problem rather than solve it.

The underlying demand uniting all of this work is straightforward: the scientific community needs evidence that the reagents used to generate biomedical findings actually measure what researchers intend them to measure. Building rigorous validation into routine practice, supported by institutions, funders, publishers, and manufacturers, is the path the new studies chart. After all, as the Primer’s title suggests, better validated reagents enable better science, and the ethical as well as the scientific stakes of getting this right could hardly be higher.

Subject of Research: Antibody validation failures in biomedical research and stakeholder-driven solutions

Article Title: Better antibodies, better science

Article References: Duchêne, J. (2026). Better antibodies, better science. PLOS Biology, 24(10), e3004001. https://doi.org/10.1371/journal.pbio.3004001

Image Credits: AI Generated

DOI: 10.1371/journal.pbio.3004001

Keywords: antibody validation, reproducibility crisis, PLOS Biology, YCharOS, knockout samples, research ethics, Delphi consensus, protein binders, artificial intelligence, biomedical reagents, animal tissue waste, stakeholder roadmap

The research team took an unusually broad, mixed-methods approach. They conducted focus groups with twelve UK-based researchers, surveyed 107 researchers and research students with confirmed institutional affiliations, and systematically analysed 785 publications linked to antibodies with objectively poor performance. The goal was to triangulate three perspectives: what researchers say they do when choosing and validating antibodies, what they actually report doing in the published literature, and what the downstream consequences are for biological samples. This between-method design allowed the authors to cross-check evidence generated from very different sources, revealing a striking gap between self-reported practice and observed behaviour.

The focus groups and survey revealed that antibody selection is driven substantially by social and pragmatic factors rather than rigorous evaluation of performance data. Previous use within the laboratory and data in the published literature were rated as the most important purchasing factors, each rated important or very important by 77.6 percent of respondents, followed by supplier reputation at 68.2 percent and citation counts at 56.1 percent. Junior researchers often simply followed supervisor direction with minimal independent evaluation. One participant admitted they had never even heard of validating antibodies until joining the study. Established habits, affordability, and the availability of free samples all shaped decisions, while the most commonly consulted form of validation data—western blot images showing a single band at the expected molecular weight—is, on its own, not robust evidence of specificity.

When it came to validation itself, the survey painted a deceptively reassuring picture. Only about one third of respondents were aware of the International Working Group for Antibody Validation (IWGAV) five-pillar framework, published in Nature Methods in 2016, which recommends genetic strategies, tagged expression, independent antibodies, immunocapture mass spectrometry, and orthogonal validation. Nevertheless, 72 percent of surveyed researchers reported having used at least one IWGAV-aligned validation method, though most relied on a single pillar and, under a stricter definition of orthogonal validation, the proportion falls to 58.9 percent. Twenty-eight percent reported no IWGAV-aligned validation at all, relying instead on basic technical controls such as isotype or secondary-only controls, and four participants reported no validation of any kind.

The published literature told a far bleaker story. The study drew on data from the YCharOS consortium, an open-science initiative that had subjected 614 commercial antibodies directed at neuroscience-relevant proteins to standardised, knockout-controlled characterisation across western blot, immunoprecipitation, and immunofluorescence. Ninety-seven of these antibodies—15.8 percent—failed in every application tested. The Leicester team then searched the literature for publications linked to these failing antibodies and analysed 785 papers. Among the 760 papers where validation status could be determined, only 120—again 15.8 percent—presented any validation evidence aligned with the IWGAV pillars. The remaining 640 papers, 84.2 percent, reported no validation whatsoever, despite employing antibodies that had failed the most definitive test available: knockout controls, in which the target protein is genetically removed and any residual antibody signal should disappear.

The sample counts extracted from these publications are sobering. Across the 640 papers without validation data, researchers reported a minimum of 8,064 animal samples—primarily mouse and rat—and 4,424 human tissue samples, along with 1,081 human primary cell samples and 400 cell lines. The authors describe these as samples at risk of waste, a deliberately careful phrase. For the global extrapolation, they reserved the term waste for the most conservative subset: samples used with antibodies that failed every tested application, lacked reported context-specific validation, and have since been withdrawn from sale by their manufacturers, rendering the associated experiments irreproducible regardless of whether validation was silently performed. Sixteen of the 35 antibodies analysed had been discontinued, and 77.1 percent carried at least one vendor-level flag such as discontinuation or a warning against use in mouse.

Applying the observed 15.8 percent failure rate to conservative estimates of 1 to 2 million distinct commercial antibodies worldwide, and multiplying by per-antibody sample consumption rates derived from the discontinued-antibody subset, the authors estimate that roughly 4 to 7 million animal samples and 6 to 11 million human tissue samples have been consumed globally without context-specific validation. These figures are explicitly lower-bound, order-of-magnitude estimates. They count only explicitly reported sample numbers, exclude unpublished research and publications predating December 2000, exclude samples used with the 19 failing antibodies that remain commercially available, and assume that antibodies which passed at least one standardised test perform reliably in every context—an assumption the authors note is almost certainly too generous, given that antibody performance varies with application, species, tissue type, and experimental conditions.

The human tissue estimates rest on the most direct evidence, because the YCharOS pipeline tested antibodies against human antigens—the same species as the tissue consumed. The animal-sample inference requires an additional step, since performance against human antigens does not automatically predict performance against orthologous proteins in other species, but convergent evidence supports the concern: many of the failing antibodies were not recommended by their own vendors for mouse use, and virtually none of the papers using them presented any validation in the experimental system employed. The authors also examined the 20 papers that did report knockout controls and found they rarely tested the same thing as the standardised characterisation; in 15 of 20, the knockout material was mouse tissue rather than the human antigen against which the antibody had failed.

Why does this happen? The researchers mapped barriers onto the COM-B behavioural model of capability, opportunity, and motivation. Survey respondents rated money (77.6 percent) and time (75.7 percent) as the most important barriers, followed by lack of supervisor support (63.6 percent) and expertise (60.7 percent). One participant characterised validation as an expensive and time-consuming hobby, and others noted that funders do not adequately account for validation costs. Motivational barriers proved equally important: when nonvalidated antibodies produce publishable results, there is little personal incentive to change, and some researchers assume that widely cited antibodies must already be adequately validated by the community. Publication bias favouring positive findings further misleads, since unexpected bands and anomalous staining patterns are often simply ignored.

Crucially, the researchers who took part strongly support structural solutions. Open data sharing of antibody performance—including negative results—received the highest support at 80.4 percent, followed by dedicated validation funding and increased visibility of negative data (both 77.6 percent), and publisher requirements for validation evidence at 70.1 percent. Participants envisioned consumer-review-style platforms where verified researchers could share antibody outcomes, alongside education at every career stage and stricter industry standards. The study’s companion Delphi consensus paper sets out a stakeholder-endorsed roadmap of actionable recommendations. The central message is that this is a well-characterised and addressable problem: demonstrating that an antibody signal is specific to its intended target is not an optional extra but a component of sound experimental design, and coordinated action by funders, publishers, manufacturers, and institutions could prevent the further, avoidable sacrifice of animal lives and human tissue donations to unreliable reagents.

Subject of Research: The ethical costs of inadequate antibody validation for animal and human tissue samples in biomedical research

Article Title: Inadequate antibody validation places substantial numbers of animal and human tissue samples at risk of waste

Article References: Biddle, M., Cooper, J., Blades, K., Ruddy, D., Krockow, E. M., & Virk, H. (2026). Inadequate antibody validation places substantial numbers of animal and human tissue samples at risk of waste. PLOS Biology, 24(10), e3003984. https://doi.org/10.1371/journal.pbio.3003984

Image Credits: AI Generated

DOI: 10.1371/journal.pbio.3003984

Keywords: antibody validation, reproducibility crisis, animal research, human tissue samples, PLOS Biology, IWGAV, knockout controls, biological sample waste, YCharOS, research ethics, biomedical reagents, open science

Antibodies are among the most widely used tools in modern biology. These Y-shaped proteins, exploited for their remarkable ability to recognize specific molecular shapes, allow researchers to detect, quantify, and isolate particular proteins within complex biological samples. Techniques ranging from Western blotting and immunohistochemistry to flow cytometry and diagnostic assays all depend on antibodies binding precisely to their intended targets. Yet the reagents sold for these purposes do not always behave as their suppliers claim. Research antibodies can fail to bind their intended targets, or they can bind additional, unintended targets, producing signals that look convincing but reflect nothing more than molecular coincidence. Studies have suggested that many commercial antibodies simply do not bind as advertised, and this lack of specificity can quietly misdirect biomedical research across virtually every field that touches proteins.

The first of the two studies, led by Virk with first author Michael Biddle, set out to measure the consequences of that failure rather than merely describe it. The team combined three strands of evidence: focus groups involving 12 researchers, a survey of 107 researchers, and an analysis of 785 publications linked to antibodies that had failed rigorous, knockout-controlled testing. Knockout validation, in which the target protein is genetically removed to confirm that the antibody’s signal disappears along with it, is considered one of the most stringent checks available, so antibodies failing such tests provide a well-grounded proxy for reagents that cannot be trusted.

The results of that analysis were sobering. Among 760 publications in which the validation status of the antibodies could be determined, only 120, or 15.8 percent, presented any validation evidence at all. This shortfall was not explained by researchers being unaware of good practice. In the survey, 72.0 percent of respondents reported having used at least one recommended validation method, a striking disconnect between what scientists say they do and what the published record shows. The papers lacking antibody validation had used a minimum of 8,064 animal samples and 4,424 human tissue samples, and when the team extrapolated from these figures, they estimated that millions of animal and human tissue samples have been consumed globally in experiments resting on inadequately validated antibodies.

“This study provides, to our knowledge, the first systematic quantification of biological sample waste attributable to the use of poorly performing antibodies without context-specific validation,” the researchers say. Biddle emphasized the scale of the problem in the starkest terms. Researchers told the team they validate their antibodies, he noted, with 72 percent reporting use of at least one recommended method, yet only 120 of the 760 papers the team could assess showed any validation evidence. The remaining 640 studies reported at least 8,064 animal and 4,424 human tissue samples used with antibodies that had failed independent testing, samples the authors describe as being at risk of waste. In the clearest cases, where the antibody has since been withdrawn from sale so that the work cannot be reproduced at all, scaling to the commercial antibody market yields a lower-bound global estimate of 4 to 7 million animal samples and 6 to 11 million human tissue samples.

Behind the numbers lies a human and ethical dimension that Virk knows personally. “Around ten years ago I discovered that data I had submitted in a grant application relied on an antibody that did not detect its intended target,” he says. “The data included staining of bronchial biopsies from patients who had given informed consent. They would not have expected their donation to be wasted. That has driven this work ever since.” That experience, a trusted reagent failing silently and rendering a patient’s contribution scientifically useless, crystallized the problem the two studies now document at scale: every unvalidated antibody is not merely a technical risk but a potential squandering of animals sacrificed, patients’ tissues donated, and public research funding spent.

Quantifying the damage was only half of the project. In the second study, Virk and first author Katherine Blades convened 32 international experts, including researchers, publishers, funders, antibody manufacturers, and institutional leaders, for a two-round Delphi consensus exercise. The Delphi method, a structured process in which participants anonymously rate and re-rate proposals across successive rounds, is designed to surface genuine agreement while filtering out the noise of group dynamics. The panel was asked to judge proposed reforms for antibody validation on two axes: whether each action would be effective, and whether it would be feasible to implement.

The consensus that emerged was remarkably concrete. The panel agreed that 15 actions, including institutional training programs, dedicated validation budgets written into grant applications, and publisher reporting requirements, were both effective and feasible for implementation by 2030. A further 15 actions were judged effective but of uncertain feasibility, marking out a second tier of reforms that could follow once the groundwork is laid. Participants identified two key barriers standing in the way: the diffuse ownership of the problem, with no single stakeholder clearly responsible for fixing it, and market incentives that fail to reward antibody quality, allowing poorly performing products to compete on price rather than performance.

“What struck us was the level of agreement once researchers, funders, publishers, institutions and manufacturers were brought into the same process,” Blades says. “The panel reached consensus on 15 actions that are both effective and achievable by 2030. Progress is held back not by disagreement about what to do, but because no single group owns the problem — so everyone waits for someone else to move first.” To translate that consensus into practice, the team has prepared separate documents for each stakeholder group, presenting the agreed recommendations alongside implementation options drawn from the panel’s qualitative feedback. These documents deliberately present options rather than prescriptions, recognizing that the optimal approach will vary across organizations, countries, and contexts.

The work is already moving from analysis to action. Since completing the studies, the team has embedded champions for better antibody practice in 14 UK research institutions, and the University of Leicester has made antibody validation training mandatory for its bioscience postgraduate researchers. The researchers are working with two national funders, the NC3Rs and Cancer Research UK, on implementation solutions, and they have built free tools that publishers can use to protect the integrity of what they publish. “Together these papers show the problem has significant impact with ethical dimensions, and is solvable,” Virk says. The message to the research community is unusually hopeful for a field accustomed to reproducibility crises: the waste is enormous, the causes are identifiable, and, for the first time, a broad coalition of stakeholders has agreed on a concrete, dated path to fixing it.

Subject of Research: Antibody validation failures and biological sample waste in biomedical research

Article Title: Poor antibody validation wastes millions of biological samples, but solutions exist

Article References: Poor antibody validation wastes millions of biological samples, but solutions exist. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: antibody validation, research reproducibility, PLOS Biology, biological sample waste, animal research, Delphi consensus, research integrity, knockout validation, University of Leicester, publishing standards, research funding, biomedical reagents