henhaobio
Sign in

PLoS BiologyCell BiologyPeer-reviewed journal article

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

Michael BiddleiD, Jemma CooperiD, Katherine BladesiD, Dominic Ruddy, Eva Maria KrockowiD, Harvinder S VirkiD

PDFDOIPreprintData
In-depth full-text digest available · Jump to digest ↓

Abstract

Lack of antibody validation by researchers frequently misdirects biomedical research, yet the ethical consequences -- particularly avoidable use of animal and human biological materials -- remain unquantified. Using focus groups (n=12), surveys (n=107), and systematic analysis of 785 publications, we examined how researchers select and validate antibodies, and quantified the downstream ethical costs. Antibody selection is substantially influenced by social factors such as previous laboratory use and peer recommendations, while systematic evaluation of performance characteristics remains limited. From a dataset of 614 antibodies subjected to rigorous characterisation using knockout controls, 97 (15.8%) failed across all tested applications; we systematically searched for publications linked to these antibodies. Within the 760 publications (those where validation status could be confirmed), only 120 (15.8%) presented any validation evidence -- despite 72.0% of surveyed researchers reporting having used at least one recommended validation method. The remaining 640 papers reported a minimum of 8,064 animal samples and 4,424 human tissue samples used with antibodies of demonstrated poor performance and without evidence of fitness for the specific experimental purpose; we describe these as samples at risk of waste. Reserving the term waste for samples used with reagents since withdrawn from sale, and therefore irreproducible, conservative extrapolation gives lower-bound, order-of-magnitude estimates of millions of animal and human tissue samples consumed globally. Researchers identified time, cost, and lack of supervisor support as primary barriers, whilst strongly supporting open data sharing, dedicated validation funding, and publisher requirements as solutions. These findings provide a first quantification of the ethical costs of inadequate antibody validation and highlight the need for coordinated stakeholder interventions to reduce avoidable biological sample waste in biomedical research.

In-depth summary

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 2 (20 Aug 2026) · CC BY 4.0

Of 760 papers using 35 antibodies that failed knockout-controlled testing in every application, 15.8% showed validation; the other 640 used at least 8,064 animal and 4,424 human tissue samples.

Poorly specific research antibodies have repeatedly misdirected biomedical work, and earlier efforts to size the problem looked at literature reliability and economic waste rather than the animal and donated human samples consumed. This study adds that count, using papers that relied on antibodies which failed standardised testing by the YCharOS (Antibody Characterisation by Open Science) consortium, and sets it against researchers' own accounts of antibody selection and validation.

Key results

  1. ShownFew papers using antibodies that failed every standardised test showed validation aligned with the International Working Group for Antibody Validation (IWGAV) five pillars, although most surveyed researchers reported using such methods.

    120 of 760 papers (15.8%) showed IWGAV-aligned validation and 640 (84.2%) showed none. Among the 120, 103 (85.8%) used knockout or knockdown, 40 (33.3%) overexpression or tagged protein, and 1 (0.8%) independent antibodies. In the survey, 77 of 107 (72.0%) reported at least one IWGAV-aligned method, and 36 of 107 (33.6%) knew the five-pillar framework. · How do researchers validate antibodies –…; How do researchers validate antibodies –…

  2. ShownPapers without validation accounted for most of the animal and human samples used with these antibodies. Within those papers, human tissue use was concentrated on antibodies since discontinued, and animal use on antibodies still sold.

    Across 766 papers with extractable counts: at least 9,443 animal, 5,015 human tissue and 1,215 human primary cell samples, plus 569 cell lines. The 640 unvalidated papers used 8,064 animal and 4,424 human tissue samples; 3,437 of those 4,424 (77.7%) involved discontinued antibodies, while 5,847 of the 8,064 animal samples (72.5%) involved antibodies still commercially available. · What is the extent of biological sample use –…

  3. SuggestedA deliberately conservative extrapolation puts global use of animal and human tissue samples without context-specific validation in the millions.

    Failure rate 97/614 (15.8%) applied to an assumed 1 to 2 million distinct commercial antibodies gives 158,000 to 316,000 failing antibodies. At about 23 animal and 35 human tissue samples per failing antibody (from 244 discontinued-antibody papers without validation, divided by 97), this yields roughly 4 to 7 million animal and 6 to 11 million human tissue samples. · What is the extent of biological sample use –…; Step 1: Per-antibody sample consumption rates; Step 2: Market size estimation; Step 3: Extrapolation calculation

  4. SuggestedAntibody choice leaned on social cues such as earlier use in the lab and published data. As validation evidence, respondents rated a single Western blot band above knockout or knockdown controls.

    Survey (n=107), share rating a factor important or very important: previous lab use and literature data 77.6% each, supplier reputation 68.2%, citations 56.1%, cost 46.7%, clonality 44.9%, free sample 40.2%. For validation data: single Western blot band at the correct weight 70.1% vs knockout/knockdown controls 63.6%. In the focus groups (n=12), junior researchers described following their supervisors' choices. · How do researchers choose antibodies –…; How do researchers choose antibodies –…

  5. SuggestedMoney and time were rated the main obstacles to validating antibodies. Respondents most favoured open data sharing, validation funding, visibility of negative results and publisher requirements, and rated undergraduate lectures lowest.

    Barriers rated important or very important: money 77.6%, time 75.7%, supervisor or principal investigator support 63.6%, expertise 60.7%, lack of incentives 37.4%. Solutions: open data sharing 80.4%, validation funding 77.6%, visibility of negative data 77.6%, publisher requirements 70.1%, third-party validation services 48.6%, undergraduate lectures 43.0%. · What are barriers to antibody validation –…; What are barriers to antibody validation –…; What are potential solutions to improve antibody…; What are potential solutions to improve antibody…

Methods, data and reproducibility

Sequential mixed methods: three virtual focus groups (reflexive thematic analysis framed by the Capability-Opportunity-Motivation-Behaviour (COM-B) model), an online Likert-scale survey, and a review of papers matched by catalogue number to antibodies failing YCharOS testing in every application, scoring IWGAV-aligned validation and explicitly reported sample counts, then extrapolating globally.

  • Virtual focus groups with UK-based researchers and research students experienced with antibodies (three groups of four, June to July 2023) · n = 12
  • Cross-sectional online survey of researchers and research students with confirmed institutional emails (October 2023); responses deposited on Zenodo · n = 107 · 10.5281/zenodo.18682587
  • YCharOS antibody characterisation panel tested with knockout controls in human cell lines; 97 failed all tested applications, and 3 neutralising-assay antibodies were then excluded · human (antigens and cell lines used in testing) · n = 614 antibodies (97 failing all applications)
  • Publications matched by catalogue number to 35 failing antibodies via CiteAb and Google Scholar, available online December 2000 to October 2025 · mouse, rat, other animal species and human (samples counted) · n = 785 papers (766 with sample counts, 760 with validation status) · 10.5281/zenodo.18682587
Methods & tools
YCharOS antibody characterisation data, CiteAb, Google Scholar, Microsoft Forms, Microsoft Teams, IWGAV five-pillar framework, COM-B model, Behaviour Change Wheel, Reflexive thematic analysis (Braun and Clarke)
Reproducibility gaps
Focus group transcripts are withheld because they may identify participants; only de-identified quotes are provided. No analysis code or scripts are mentioned.
Limitations (6)
  • Focus groups were small (n=12) and mainly junior researchers, limiting generalisability to senior investigators; both focus groups and the survey are open to selection, social desirability and recall bias. · stated by the authors
  • The literature review sees only reported validation, so validation done but not published or cited is missed; papers using antibodies that passed standardised testing were not analysed for comparison. · stated by the authors
  • Standardised testing used human antigens, so applying it to animal samples is an indirect inference. The authors rely on vendor flags (27 of 35 antibodies discontinued or not recommended for mouse) and note the added uncertainty. · stated by the authors
  • The global estimate assumes the 614-antibody panel's failure rate holds market-wide and rests on an uncertain count of 1 to 2 million distinct antibodies; discontinuation may reflect commercial rather than quality reasons, and commercial status shifted during the study. · stated by the authors
  • Only 35 of the 94 eligible failing antibodies had linked publications, from a panel aimed at neuroscience-relevant proteins, and the survey and focus groups used snowball sampling, so neither the papers nor the respondents are random samples. · noted in this summary
  • The 72.0% self-reported vs 15.8% published validation contrast spans different groups: survey respondents were sampled separately from, and not known to include, the 760 papers' authors, and the papers cover only antibodies that failed testing, so it is not one population's gap between stated and actual practice. · noted in this summary

Read this if you use antibodies on animal or human tissue, set reporting or funding policy on reagent validation, or want an example of checking survey self-reports against audited publication practice.

Citations
3
FWCI
22.60
Field percentile
top 1%

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISBlades '26Ayoubi '23Bolognesi '21Pace '22Mazoit '25Schoustra '18
72% semantic matchScientific Communication & Education
Katherine Blades, Michael Biddle, Robert Froud+2 more·bioRxiv·Aug 20, 2026