henhaobio
Sign in

eLifeMolecular BiologyPeer-reviewed journal article

Scaling of an antibody validation procedure enables quantification of antibody performance in major research applications

Riham AyoubiiD, Joël RyaniD, Michael BiddleiD, Walaa AlshafieiD, Maryam FotouhiiD, Sara González BolívariD, Vera Ruíz MoleóniD, Peter EckmanniD, Donovan WorralliD, Ian McDowelliD, Kathleen SoutherniD, Wolfgang ReintschiD, Thomas M. DurcaniD, Claire M. BrowniD, Anita E. BandrowskiiD, Harvinder S VirkiD, A.M. EdwardsiD, Peter Scottt McPhersoniD, Carl LaflammeiD

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

Abstract

Antibodies are critical reagents to detect and characterize proteins. It is commonly understood that many commercial antibodies do not recognize their intended targets, but information on the scope of the problem remains largely anecdotal, and as such, feasibility of the goal of at least one potent and specific antibody targeting each protein in a proteome cannot be assessed. Focusing on antibodies for human proteins, we have scaled a standardized characterization approach using parental and knockout cell lines (Laflamme et al., 2019) to assess the performance of 614 commercial antibodies for 65 neuroscience-related proteins. Side-by-side comparisons of all antibodies against each target, obtained from multiple commercial partners, demonstrates that: i) more than 50% of all antibodies failed in one or more tests, ii) yet, [~]50-75% of the protein set was covered by at least one high-performing antibody, depending on application, suggesting that coverage of human proteins by commercial antibodies is significant; and iii) recombinant antibodies performed better than monoclonal or polyclonal antibodies. The hundreds of underperforming antibodies identified in this study were found to have been used in a large number of published articles, which should raise alarm. Encouragingly, more than half of the underperforming commercial antibodies were reassessed by the manufacturers, and many had alterations to their recommended usage or were removed from the market. This first such study helps demonstrate the scale of the antibody specificity problem but also suggests an efficient strategy toward achieving coverage of the human proteome; mine the existing commercial antibody repertoire, and use the data to focus new renewable antibody generation efforts.

In-depth summary

Claude Opus 5.5 · max effort · 8 Nov 2026 · from the full text of version 6 (17 Aug 2023) · CC BY 4.0

Over half of 614 commercial antibodies for 65 human neuroscience proteins failed a knockout-controlled assay, yet most targets had a well-performing antibody; recombinants performed best.

About 1.6 million commercial antibodies target ~96% of human proteins, but how many miss their intended targets has been mostly anecdotal and validation practices differ in rigor, so progress toward a reliable antibody per protein could not be gauged. A partnership including 10 manufacturers applied standardized knockout-based protocols at scale, yielding open data on failure rates, renewable-antibody coverage and the value of vendor validation methods.

Key results

  1. ShownUnder uniform parental-versus-knockout testing, over half of the antibodies failed at least one of Western blot (WB), immunoprecipitation (IP) or immunofluorescence (IF), yet most targets had a well-performing antibody, with IF coverage weakest.

    >50% of 614 antibodies failed one or more tests; ~50-75% of the 65 targets had a high-performing antibody depending on application. A successful renewable (monoclonal or recombinant) antibody existed for 77% of targets in WB (50/65), 75% in IP (49/65) and 54% in IF (30/56); ~40% of targets lacked any successful IF antibody. · Abstract; Coverage of human proteins by renewable…; Discussion

  2. ShownRecombinant antibodies achieved the highest success rates in every application, ahead of hybridoma monoclonal and polyclonal antibodies, supporting a shift of new antibody generation toward recombinant formats.

    Pool of 258 polyclonal, 165 monoclonal and 191 recombinant antibodies. Success (polyclonal/monoclonal/recombinant): WB 27%/41%/67%; IP 39%/32%/54%; IF, 529 antibodies against intracellular targets, 22%/31%/48%. · Recombinant antibody performance; Discussion

  3. ShownVendor recommendations were an unreliable guide to knockout-confirmed performance, most clearly for IF, and IP was under-recommended; IF antibodies that vendors had validated with genetic (knockout or knockdown) controls held up far better than orthogonally validated ones.

    WB: of 578 vendor-recommended antibodies, 44% succeeded, 35% were specific but non-selective, 21% failed. IP: 58% of 143 recommended vs 37% of 471 non-recommended enriched their target. IF: 39% of 293 recommended succeeded; 38% of orthogonally validated vs 16 of 20 (80%) genetically validated were confirmed (WB target detection: 80% vs 89%). · Optimizing an antibody characterization strategy

  4. ShownAntibodies that failed here have been used extensively in the literature, and most IF papers relying on a failing antibody lacked validation data, suggesting a sizeable fraction of published antibody-based figures may be unreliable.

    CiteAb: share of publications using a failing antibody was 31% for WB (2010 papers, 180 antibodies), 35% for IP (105 papers, 41 antibodies) and 22% for IF (548 papers, 80 antibodies); 88% of 112 checked IF papers with a failing antibody showed no validation data; ~12 such papers per target on average; authors extrapolate to 20-30% of figures. · Antibodies and reproducible science; Discussion; Data analysis

  5. ShownSharing results with suppliers prompted catalogue corrections: manufacturers reassessed more than half of the underperforming antibodies, withdrawing some and revising usage recommendations for others.

    Of 409 antibodies whose data conflicted with supplier recommendations, participating companies reported withdrawing 73 from the market and changing recommendations for 153; one of the 4 IF antibodies that vendors had genetically validated but that failed here was already withdrawn. · Abstract; Optimizing an antibody characterization strategy; Discussion; Data analysis

Methods, data and reproducibility

Each vendor-supplied antibody was tested side by side on a parental cell line, chosen via Cancer Dependency Map (DepMap) RNA levels, and its CRISPR knockout by Western blot, immunoprecipitation (blot readout) and mosaic immunofluorescence; outcomes pooled from 65 target reports were compared by clonality, vendor claims, validation method and citations.

  • Antibody Characterization through Open Science (YCharOS) reports, one per target, with parental vs knockout WB, IP and IF data, peer reviewed by advisors · human (cell lines and target proteins) · n = 65 reports
  • Commercial antibodies selected and supplied by partner manufacturers (average 9.5 per target from an average of 5 vendors); IF only for intracellular targets · human (target proteins) · n = 614 (258 polyclonal, 165 monoclonal, 191 recombinant) · Research Resource Identifier (RRID) per antibody, e.g. AB_2037651
  • Funder-nominated neuroscience-related targets from Alzheimer's, amyotrophic lateral sclerosis (ALS), Parkinson's and industry lists; 56 predicted intracellular, 9 secreted · human · n = 65 (32 Alzheimer's, 22 ALS, 5 Parkinson's, 6 industry)
  • Parental and CRISPR knockout cell line pairs; parental lines picked from DepMap Expression 22Q1 RNA data above 2.5 log2(TPM+1); knockouts sourced from partners or made in-house · human (DepMap cancer cell lines) · n = 8 backgrounds for 62 of 65 targets; >1800 DepMap lines searched
  • Antibody Registry counts of commercial antibodies per human protein, split into polyclonal and renewable, used to estimate proteome coverage · human · n = ~1.6 million antibodies · SCR_006397
  • CiteAb records of publications using the tested antibodies, by application (searched November 2022 to March 2023) · n = 2010 WB, 105 IP and 548 IF publications
  • Manually verified CiteAb immunofluorescence/immunocytochemistry papers (Homo sapiens reactivity) using failing IF antibodies, scored for validation data · human · n = 112 publications
  • NIH protein database publication counts per target, used to gauge how well studied each protein is (searched May 4, 2023) · human · n = 65 targets (15 with >500 publications)
Methods & tools
Parental vs isogenic CRISPR-Cas9 knockout cell line comparison, Western blot, Immunoprecipitation with Western blot readout, Mosaic immunofluorescence of dye-labelled parental and knockout cells, Cancer Dependency Map (DepMap) Expression 22Q1, UniProt, Antibody Registry / RRID Portal, CiteAb, NIH protein database, McNemar test (chi-square, 1 degree of freedom), ZENODO (YCharOS community), F1000 publication platform
Reagents & environment
Parental and isogenic CRISPR-Cas9 knockout (KO) human cell lines (8 common backgrounds for 62 of 65 targets; KO lines from industry partners or made in-house); 614 commercial antibodies, each with an RRID (e.g., AB_2037651).
Experimental conditions
No instruments named; WB on parental vs KO lysates (media for secreted targets), IP from non-denaturing extracts read by WB, IF on dye-labelled parental/KO mosaics in one field; target RNA above 2.5 log2(TPM+1) to suit 1-50 nM-affinity antibodies.
Reproducibility gaps
No analysis code or scripts released; per-antibody outcomes are spread across 65 separate ZENODO reports, with cell lines listed in Table 1; imaging hardware and step-by-step protocols are not given in the main text, which cites prior work.
Limitations (6)
  • Recombinant antibodies are newer reagents, so their higher success rates could partly reflect more extensive internal characterization by suppliers rather than the recombinant format itself. · stated by the authors
  • Every antibody was run with one universal protocol per application; target-optimized protocol variants used by individual labs can strongly change performance and may yield outcomes different from those reported. · stated by the authors
  • Coverage estimates are probably conservative, because the participating manufacturers account for only about 27% of worldwide antibody production. · stated by the authors
  • Knockout-based validation costs ~$25,000 USD per protein and an estimated $50 million USD proteome-wide (excluding antibodies and knockout lines), and scarce knockout lines in expressing cells remain a main barrier. · stated by the authors
  • Targets were funder-nominated, mostly linked to Alzheimer's, ALS or Parkinson's disease, and manufacturers chose which antibodies to submit (prioritizing renewables), so failure and coverage rates may not generalize proteome-wide. · noted in this summary
  • Group comparisons (clonality, vendor validation method) are pooled percentages without confidence intervals or per-target adjustment, and the genetic IF validation arm holds only 20 antibodies, so effect sizes are imprecise. · noted in this summary

Read this if you select, validate or produce commercial antibodies and want knockout-based success rates for WB, IP and IF broken down by clonality and vendor validation strategy.

Citations
68
FWCI
11.21
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
THISChernigovsk '25Biddle '26Blades '26Fredolini '18Wang '20Fantini '17
56% semantic matchImmunology
Maria Chernigovskaya, Khang Lê Quý, Maria Stensland+16 more·Cell Systems·Nov 1, 2025· 2 citations