assay · intermediate · partly written
Western blot
A Western blot sorts the proteins in a lysate by size on a gel, moves them to a membrane and uses an antibody to show how much of one protein is there.
Also called: Western blotting, immunoblot
The short version
A Western blot asks one narrow question: is there less or more of a protein of this apparent size that this antibody recognises? The number is only as good as the antibody (does it see only the target?) and the signal (is the band inside the range where brighter means more?). Both fail silently, and neither shows in a figure. 12
Reported, not yet replicated
How it works, step by step more
1. Lyse the cells in detergent buffer with inhibitors. 2. SDS-PAGE: SDS coats proteins with uniform negative charge, so a current drags them through a gel mesh and they sort mainly by size. 3. Transfer to a membrane. Towbin, Staehelin and Gordon (1979) wanted proteins in the open where antibodies could reach them; their method detected as little as 100 pg of protein, and the abstract says transfer from SDS gels kept the band pattern but was not quantitative. The name "Western" comes from Burnette 1981, whose title is "“Western blotting”: electrophoretic transfer of proteins from sodium dodecyl sulfate-polyacrylamide gels to unmodified nitrocellulose"; Renart, Reiser and Stark had published transfer to a reactive paper with antiserum detection two years earlier, in 1979. Both papers are cited here from their catalogue records only, so this is the naming and the dates and not an account of either method. 4. Block the empty membrane so antibody does not stick everywhere. 5. Primary antibody binds the target epitope; a secondary antibody carrying an enzyme or fluorophore binds the primary. Detection is by light (enzyme plus ECL substrate) or fluorescence. 6. Read band intensity at the expected size.
The deepest level that matters is antibody-epitope recognition (step 5); the chemistry of the light reaction does not change what you conclude. 3456
Reported, not yet replicated
Where it breaks more
Non-specific antibody. Antibodies differ by supplier and condition: one study validated five commercial ubiquitin antibodies (they differed in detecting free versus conjugated ubiquitin) and six ISG15 antibodies, where most gave different semi-quantitative results. Uhlen and colleagues list five validation pillars: genetic knockout or knockdown (their siRNA example cut the blot signal by 90% and 98%), an orthogonal method, a second antibody against another epitope, a tagged protein, and immunocapture plus mass spectrometry. Genetic strategies cannot be used for human tissue samples and body fluids.
No accepted standard. Pillai-Kastoori and colleagues cite the lack of clear, accepted standards for antibody validation and reporting, and advise confirming the effect by a complementary or orthogonal method. Gilda proposes a minimal reporting standard (antibody catalogue number, loaded amount, blocking, detection).
Signal saturation. LI-COR (a vendor) describes two types, membrane and imager saturation, and suggests a serial dilution of 8 to 12 two-fold points to find the linear range for target and control, then working mid-range.
Bad loading control. In rat brain homogenates a total protein stain was much less variable than tubulin, actin or GAPDH (Kirshner 2018).
Not found: how often blot-based conclusions are wrong. 17829
Reported, not yet replicated
How it gets abused deep
Image duplication is the documented problem. Bik, Casadevall and Fang screened 20,621 papers containing the term "Western blot" from 40 journals, 1995 to 2014, and found 3.8% with problematic figures, at least half with features suggesting deliberate manipulation; the rate rose over the decade and varied by journal. Their screen covered all image types, not blots alone. About 40% of the sample (8,138 papers) was one journal (PLoS One, 2013 to 2014), so it is not a random sample of biomedicine, and the categories run from simple duplication (which may be honest error) to duplication with alteration.
Cropping, over-exposure and choosing the loading control that gives the wanted ratio are plausible abuses that follow from the linear-range and normaliser problems, but no source opened measures them. 10
Reported, not yet replicated
Cost and throughput deep
The only price found is $175 to $250 per assay at one academic core (menu PDF created 2025-01-27), with the unit unstated and reagents and antibody probably extra; imager time was $10 per hour self-service. No throughput figure survived verification, so the throughput class here is an estimate (see the throughput field). 11
Reported, not yet replicated
Question answered
Is there less (or more) of a protein of this apparent size that this antibody recognises? It is the readout that converts "the cells respond" into "the target protein level went down", and in a degrader programme it defines potency. It says nothing about why the level changed. 4
Reported, not yet replicated
Readout physical
Not measured
Why a band shrank: degradation, less transcription, less translation or dead cells all lower it, so degrader papers add proteasome and ligase-dependence experiments. That the band is the target: size is only a proxy for identity. Absolute amount, unless a standard curve is run (results are semi-quantitative). Single-cell or spatial information, because a lysate is a population average. 4128
Reported, not yet replicated
Setting
Where in cascade
Sample prep
Cells are lysed in detergent buffer with protease and phosphatase inhibitors (the ARV-110 paper used RIPA) so that proteins are freed and not chewed up. That paper loaded 2 ug of VCaP lysate and 5 ug of other lines. The lysate is then separated by SDS-PAGE, transferred to a membrane (nitrocellulose or PVDF), blocked with milk or BSA, and probed with primary then secondary antibody. 412
Reported, not yet replicated
Conditions
The antibody (supplier, clone, epitope) and the signal range decide the answer. Above the linear range extra protein stops giving extra signal, through membrane saturation (overloading) or imager saturation, so strong bands are underestimated. The choice of loading control matters: housekeeping proteins can vary with treatment or saturate. 291
Reported, not yet replicated
Throughput
Our estimate, not a sourced figure: a blot reads a handful of lanes per gel, so a dose-response with a few concentrations in duplicate fits on a gel or two, while a library of thousands does not. This is why it sits late in the cascade and only sees compounds that survived cheaper screens. No throughput number was retained from the sources. 11
Our inference
Cost
Per sample
175 to 250 USD per assay (unit not stated on the menu) (UNMC Bioassay Core service menu, academic core rate; reagents and antibody likely extra but not stated. Imager time is listed separately at $10 per hour (iBright FL1000, self-service). This is an academic core-facility recharge rate, not a commercial or contract-lab price; no commercial price was found.; as of 2025-01-27)11
Engineerable variants
Capillary immunoassay (Simple Western, Jess, Leo)
Gel and transfer are replaced by a capillary. Vendor claims of lower sample need and less hands-on time were not confirmed here. 12
Our inference
Fluorescent two-colour detection with total-protein stain
Target and normaliser are read on the same membrane. In rat brain homogenates a total protein stain was much less variable across samples than three housekeeping proteins (tubulin, actin, GAPDH). 92
Reported, not yet replicated
Multi-gel calibration curve
A 2026 method uses log-logistic regression to calibrate signal against protein amount, to raise the sample count with proper statistics. Read as an abstract only. 13
Reported, not yet replicated
HiBiT tag in place of an antibody
The endogenous protein is tagged with an 11-amino-acid HiBiT tag by CRISPR, and light output tracks level in plate format, giving rate, Dmax and DC50. The abstract says profiling target degradation in a high-throughput fashion remains highly challenging. The tag is itself a change to the protein. 14
Reported, not yet replicated
Use cases
ARV-110 (bavdegalutamide), Arvinas
Androgen receptor degrader. The paper reports a DC50 of 1 nmol/L in LNCaP and VCaP cells with over 95% of AR degraded (24 h), and only 10% of AR remaining in VCaP at 10 nmol/L after 4 hours, with AR polyubiquitination and proteasome and ligase dependence shown. Methods describe Western blotting with AR clone D6F11 and GAPDH as loading control, and an ubiquitin pull-down read by blot; label-free mass spectrometry proteomics at 10 nM for 8 h was an orthogonal selectivity check. The report infers, not from figure labels, that the dose-response panels are blots. It was the first PROTAC protein degrader to enter human clinical trials (NCT03888612). This is the company's own account; clinical results were not read. 4
Reported, not yet replicated
ML role
Does
No source opened describes machine learning on Western blot data. The output is an image reduced by densitometry to one number per lane. 12
Our inference
Cannot
A model cannot repair what the figure hides: whether the antibody sees only the target, and whether the band sat inside the linear range. A blot is a noisy edge in a knowledge graph (compound X lowers protein Y), and the antibody identity and normaliser decide whether that edge is believable, so records should keep both. 12
Our inference
Teach this
In one sentence
A Western blot sorts proteins by size, then uses an antibody to show whether there is more or less of one of them, and it cannot say why. 124
Reported, not yet replicated
The five-minute version
Lyse cells, run the lysate through a gel so proteins sort by size, transfer them to a membrane, and probe with a primary antibody and a labelled secondary antibody. The band at the right size is the target, if the antibody is specific. Brightness means amount only inside the linear range, and is divided by a loading control to correct for how much lysate was loaded. In a degrader programme such as ARV-110 the blot defines potency (DC50 and Dmax), but a fainter band alone does not separate degradation from lower expression or cell death, so proteasome and ligase-dependence experiments are added. 421
Reported, not yet replicated
What a listener will ask
Does a smaller band prove the protein is being degraded?
No. Degradation, less transcription, less translation or dead cells all lower the band; degrader papers add proteasome and ligase-dependence experiments for this reason.
How do you know the band is the right protein?
Size is only a proxy. Validation uses a knockout or knockdown that should abolish the signal, a second antibody against another epitope, an orthogonal method, and similar checks.
Is the number quantitative?
Semi-quantitative. It holds only inside the linear range, so labs run a dilution series for target and control and work mid-range.
The common wrong idea
A Western blot band intensity is a direct measure of how much protein there is.
Band intensity tracks amount only inside the linear range, for an antibody that sees only the target, divided by a normaliser that does not itself change.
The one citation to hand over: 4
Still open here
- How many samples does a classic blot run in practice, and how long does one run take from lysate to result? The report kept no throughput or time figure (the search-summary figures were unverified and removed), so throughput_class is our estimate and time_per_run is left out.
- What does a blot cost per sample with antibody and reagents included? The only price found ($175 to $250 per assay at one academic core) does not state its unit or whether reagents are extra, and no per-campaign range has any basis.
- How often are blot-based conclusions wrong? No source opened measures it, nor how often cropping, over-exposure or choice of loading control is abused.
- Capillary automated immunoassays (Simple Western and similar): the vendor claims about sample need and hands-on time and any list price are unconfirmed (page not opened).
- Does a total-protein stain widen the linear range compared with a housekeeping protein? Kirshner 2018 shows it varies less in rat brain, but the wider-range claim was not shown in the abstract.
- No before-and-after (breakthrough watch) and no capital or funding entries are written: the report does not support them. Data availability for blot data (public repositories, vendors) was not researched.
- Machine-learning use on blot images and densitometry was not covered by any source opened.
- The ARV-110 clinical results (NCT03888612) were not read, so what blot-defined potency meant in patients is unknown here.
Not yet written: Time per run, Before after, Capital, Data availability.
Linked from
Sources
- Gilda et al.. Antibody validation by Western blot: ubiquitin and ISG15 antibodies (Gilda et al., 2015). PLoS ONE (abstract via Europe PMC) (2015). paper; opened 2026-11-17 all records citing this
- LI-COR: Western blot linear range determination. LI-COR Biosciences. vendor-page, not peer reviewed; opened 2026-11-17 all records citing this
- Towbin, Staehelin, Gordon. Electrophoretic transfer of proteins from polyacrylamide gels to nitrocellulose sheets (Towbin, Staehelin, Gordon, 1979). PNAS (abstract via Europe PMC) (1979). paper; opened 2026-11-17 all records citing this
- Snyder et al.. Snyder et al. 2025, ARV-110 (bavdegalutamide) preclinical characterisation, Mol Cancer Ther. Molecular Cancer Therapeutics (PMC) (2025). paper; opened 2026-11-17 all records citing this
- Burnette WN. "Western blotting": electrophoretic transfer of proteins from sodium dodecyl sulfate-polyacrylamide gels to unmodified nitrocellulose and radiographic detection with antibody and radioiodinated protein A. Analytical Biochemistry 112(2):195-203 (1981). paper; opened 2026-11-19 all records citing this
- Renart J, Reiser J, Stark GR. Transfer of proteins from gels to diazobenzyloxymethyl-paper and detection with antisera: a method for studying antibody specificity and antigen structure. Proceedings of the National Academy of Sciences USA 76(7):3116-3120 (1979). paper; opened 2026-11-19 all records citing this
- Uhlen et al.. A proposal for validation of antibodies (Uhlen et al., 2016). Nature Methods (PMC) (2016). paper; opened 2026-11-17 all records citing this
- Pillai-Kastoori et al.. Antibody validation for Western blot: by the user, for the user (Pillai-Kastoori et al., 2020). Journal of Biological Chemistry (abstract via Europe PMC) (2020). paper; opened 2026-11-17 all records citing this
- Kirshner et al.. Total protein stain versus housekeeping proteins for Western blot normalisation in rat brain (Kirshner et al., 2018). Europe PMC (abstract) (2018). paper; opened 2026-11-17 all records citing this
- Bik, Casadevall, Fang. The prevalence of inappropriate image duplication in biomedical research publications (Bik, Casadevall, Fang, 2016). mBio (PMC) (2016). paper; opened 2026-11-17 all records citing this
- Western blot (Wikipedia). Wikipedia (2026). web, not peer reviewed; opened 2026-11-17 all records citing this
- Cheng et al.. Calibration-curve method for quantitative Western blotting (Cheng et al., 2026). Europe PMC (abstract) (2026). paper; opened 2026-11-17 all records citing this
- Riching et al.. HiBiT tagging to profile target degradation (Riching et al., 2020, JoVE). Journal of Visualized Experiments (abstract via Europe PMC) (2020). paper; opened 2026-11-17 all records citing this
- Amit's dictation of 2026-11-17: what the atlas must contain. Kept in the atlas repository; not published.
Where it sits on the map
History of this record (5)
This record has been changed 5 times, on the dates below. The note kept against each change is written for the people who maintain the atlas, in their own shorthand, so it is not shown here; it is in the record itself.
- Throughput class
- low
- Emerging
- no
- Status
- partly written
- Last updated
- 2026-11-19
- Record
- content/assays/western-blot.json