assay · intermediate · partly written
Surface plasmon resonance (SPR)
Surface plasmon resonance measures, in real time and without labels, how a compound or protein binds a protein fixed to a gold surface, giving on and off rates as well as affinity.
Also called: SPR, Biacore
The short version
SPR does not see the compound. It sees the refractive index within about 300 nm of a gold film rise when mass lands on the immobilised protein. A small molecule is a tiny mass, so every measurement is a fight against bulk-solvent signal. In return it is the one cascade step that gives rates, not just a yes or no. 1
Reported, not yet replicated
How it works more
1. A thin gold film on glass is lit at an angle. At one angle light couples into surface plasmons (electron waves in the metal) and the reflected intensity dips. 2. The plasmon's field leaks about 300 nm into the buffer, so the dip angle depends on the refractive index there. 3. The ligand is fixed to the surface; the analyte flows past. 4. Binding adds mass, raises the refractive index and shifts the dip angle (1 RU is 10^-4 degree, about 1 pg per mm^2). 5. The sensorgram has a baseline, association (rate set by kon and concentration), steady state, then buffer-only dissociation (rate koff). KD = koff/kon. 6. A reference channel is subtracted.
Source of the physics is a 2001 book chapter describing 1999-era instruments; the physics holds, instruments have improved. Biacore was founded in 1984 as Pharmacia Biosensor, released its first instrument in 1990 and was sold in June 2006 for $390 million (Wikipedia, secondary). The problem it solved was measuring binding in real time without labelling either partner. 12
Reported, not yet replicated
The small-molecule problem and rates more
Signal scales with analyte mass over ligand mass. For an analyte of Mr about 200 on a ligand of Mr about 40,000, about 10,000 RU of active ligand is needed to see 50 RU (the proportion checks out: 10,000 x 200 / 40,000 = 50). Refractive-index effects then become as large as the binding signal.
On rates: the complex lifetime is about 1/koff (our reading of the definition). SPR resolves koff from roughly 10^-5 to about 1 per s, and kon faster than about 10^6 per M per s is hard to measure. A tight KD alone does not say whether a complex is long-lived or short-lived, which is the point of the kinetics. 1
Reported, not yet replicated
Where it breaks more
DMSO bulk mismatch: compounds sit in DMSO, a 0.5% error at 5% DMSO shifts the response about 800 RU; check with a solvent-correction curve and matched dilutions.
Non-specific and aggregating compounds: at fragment-screening concentrations, sticking to the chip gives false positives; the abstract of Navratilova and Hopkins supports multiple referencing plus ligand-efficiency filtering to cut them (the detergent and competition details were not confirmed).
Mass transport and rebinding: fast binding at high ligand density outruns diffusion, so measured rates are too slow. Check by repeating at three or more ligand densities, at least two-fold apart, and at higher flow rate; if rates change with density, lower the surface density.
Dead or altered ligand: immobilisation can damage activity, so a negative result needs a positive control.
Kinetics are easy to generate and hard to trust: fit quality does not prove the model. Binding is not function. 341
Reported, not yet replicated
A per-compound illustration deep
If the vendor throughput held, $110 per hour x 24 hours / 2,000 compounds is about $1.30 per compound in instrument time alone. That excludes chips, protein, plate preparation and failures, and no independent throughput or per-compound price was found. Treat it as an illustration, not a price. 56
Our inference
Reading the clocks: what koff and kon mean as times more
The sensorgram gives two clocks. Kd = koff/kon; the lifetime of one complex is 1/koff. The 'on' clock has a physical ceiling: two proteins finding each other by diffusion cannot associate faster than about 1e9 to 1e10 per molar per second for uniform spheres, and a reactive-patch model gives 1e4 to 1e6, which the authors say matches many measured complexes. SPR reads koff from roughly 1e-5 to about 1 per s and reaches 99% of equilibrium in about 4.6/koff seconds. See also the concept record on binding kinetics. 17
Reported, not yet replicated
What a picosecond, a microsecond and a second mean for a bond more
Arithmetic from Kd = koff/kon with kon at the diffusion ceiling of 1e9 per molar per second (our calculation, not a measurement). koff 1e12 per s (a lifetime of 1 picosecond) gives Kd 1,000 M: a collision, not a bond, so 'very strong but picoseconds' cannot be one binding event. koff 1e6 per s (1 microsecond) gives Kd 1 mM: weak, relevant only at very high concentration. koff 1 per s (1 second) gives Kd 1 nM: signalling-grade binding. koff 1e-3 per s (about 17 minutes) gives 1 pM, and koff 1e-5 per s (about 28 hours) gives 10 fM. With kon at 1e6 each Kd is 1,000-fold larger. 7
Our inference
Measured cases deep
p85 SH2 domains binding IRS-1 phosphopeptides had kon 3e7 to 4e8 per molar per second, koff 0.11 to 0.19 per s (lifetimes about 5 to 9 seconds) and Kd 0.3 to 3 nM, read from the abstract: high affinity from a very fast on-rate, with rapid exchange. Across 144 structurally characterised complexes, Kd spans about 1e-5 to 1e-14 M (abstract only). Because measured kon spans about 1e3 to 1e10, 'koff does the sorting' is a rule of thumb for the typical kon band, not a law. 897
Reported, not yet replicated
Question answered
Does this compound physically bind this purified, immobilised protein, and how fast does it go on and come off? It is the one cascade step that returns rates (kon, koff) and so a dissociation constant (KD = koff/kon), not just a yes or no or an IC50. 1
Reported, not yet replicated
Readout physical
The shift in the angle at which reflected light dips, which depends on the refractive index within about 300 nm of a gold film. Binding adds mass at the surface, raises the refractive index and shifts the angle. One response unit (RU) is a shift of 10^-4 degree, roughly 1 pg of protein per mm^2. The plot of RU against time is the sensorgram. 1
Established
Not measured
Function, and anything in cells. A binder in this assay can be inert or bind at an irrelevant site, which is why SPR sits before cell target-engagement assays. It does not see the compound as such, only mass landing on the surface, and complexes that last picoseconds would look like no binding. 110
Reported, not yet replicated
Setting
Where in cascade
Sample prep
The protein (ligand) is fixed to the surface, often in a 100 to 200 nm dextran gel, by amine coupling (covalent) or by capture through a tag or biotin. The partner (analyte) flows past in buffer, in a flow cell of 20 to 60 nL. Compounds sit in DMSO, so a solvent-correction cycle is run and re-run every 20 to 30 samples. 13
Reported, not yet replicated
Conditions
Ligand density: low for kinetics (it cuts mass-transport artefacts), high for small molecules (they need signal). A reference channel and blank injections are subtracted, because anything that changes the buffer's refractive index also gives signal. DMSO is the largest trap: a 0.5% error at 5% DMSO shifts the response by about 800 RU, against binding signals of tens of RU. 13
Reported, not yet replicated
Throughput
Not confirmed independently. A CRO page claims a Biacore 8K service with 8 channels and more than 2,000 compounds per day (a vendor claim, which would be high throughput if it held), and a vendor article claims an array SPR system runs 384 ligands in one run. The 2001 Oxford chapter says SPR is not well suited to high-throughput work or to analytes under Mr 1000, though instruments have improved since 1999. Class is set to medium as a cautious estimate. 6111
Our inference
Time per run
Time to 99% of equilibrium is about 4.6/koff, so a koff of 0.01 per s needs about 460 s per injection (our arithmetic from the Oxford chapter). SPR resolves koff from roughly 10^-5 to about 1 per s, so it sees complexes that last seconds to hours. No whole-run time was found. 1
Our inference
Cost
Per sample
260 USD per sample, full service, including 3 replicates (Scripps Biacore S200 full-service rate; excludes the $500 setup. Academic core rate, page says subject to change. Duke internal-user instrument time is separate, $110 per hour independent to $342 per hour for prep and QC. This is an academic core-facility recharge rate, not a commercial or contract-lab price; no commercial price was found.; as of 2026)125
Per campaign
from 760 USD for one sample with optimisation, consumables extra (Scripps: first sample with optimisation $760 plus consumables. This is a floor for a single sample, not a screening-campaign price; no campaign-scale price was found. This is an academic core-facility recharge rate, not a commercial or contract-lab price; no commercial price was found.; as of 2026)12
Engineerable variants
Immobilisation chemistry
Amine coupling is covalent and can damage the protein; capture (tag or biotin) allows a fresh surface each cycle. 1
Reported, not yet replicated
Double referencing and solvent correction
Blank buffer injections plus a reference channel remove drift and bulk effects; a DMSO solvent-correction cycle costs run time. 34
Reported, not yet replicated
Bio-layer interferometry (BLI)
A fibre tip is dipped into samples with no flow. A competing vendor says off-rates are measurable only up to 0.1 per s versus 1 per s for SPR and 10 per s for grating-coupled interferometry, and that the sensor is orders of magnitude less sensitive; vendor-only wording. One university core lists BLI (Octet) at $300 setup plus $160 per sample, run in triplicate (page undated). 1314
Reported, not yet replicated
Grating-coupled interferometry (GCI)
Reads a waveguide phase shift rather than a dip angle, with a claimed faster off-rate range. Vendor-only comparison. 13
Reported, not yet replicated
Array SPR
Hundreds of spots per run (384 ligands in one run, vendor article), giving throughput for antibodies; suitability for small molecules is not shown in the sources. 11
Reported, not yet replicated
Use cases
CACHE Challenge #1: LRRK2 WDR domain
Twenty-three teams each predicted up to 100 compounds for the WDR domain of LRRK2, a Parkinson's disease protein with no known ligands. Of 1,955 molecules, 73 bound in an SPR assay with KD under 150 uM (about 3.7%, our arithmetic); after a second round of analogs, seven chemically diverse series were confirmed in two orthogonal assays at 18 to 140 uM. SPR was the arbiter of computation, and 150 uM is a loose definition of binder. Abstract and an NSF record only; full text not opened. 1015
Reported, not yet replicated
ML role
Does
Cannot
SPR data cannot say whether a binder is functional, and kinetic fits can look convincing while being wrong ("it is deceptively easy to generate kinetic data"), so a model trained on constants inherits those artefacts. The authors' conclusions on machine learning versus docking in CACHE #1 were not confirmed here. 110
Our inference
Data availability
Reaction Biology SPR service cro
A contract SPR service on a Biacore 8K system. The page claims more than 2,000 compounds per day. 6
Caveats. No price shown; page undated; throughput is a vendor claim.
Reported, not yet replicated
Scripps Biophysics and Biochemistry Core, full service cro
Full-service SPR on a Biacore S200: $500 setup plus $260 per sample including 3 replicates; first sample with optimisation $760 plus consumables. 12
Caveats. Academic core rate, subject to change, and not an industry list price.
Price. 260 to 760 USD (per sample, to first optimised sample) (Scripps core rate card, 2026 page, excluding setup and consumables; as of 2026)12
Reported, not yet replicated
Duke Biomolecular Interaction Analysis core, self-service in-house
Biacore T200 instrument time at internal-user rates: $110 per hour independent, $229 per hour operator-run, $342 per hour for prep and QC; sample prep $115 and data analysis $80; consumables not included. 5
Caveats. Internal-user rates; page undated.
Price. 110 to 342 USD per hour (Duke core, Biacore T200, internal-user rates, consumables excluded; as of page undated)5
Reported, not yet replicated
Scripps Biophysics and Biochemistry Core, BLI (Octet) cro
Bio-layer interferometry: $300 setup plus $160 per sample, triplicate. 14
Caveats. Academic core rate, subject to change; the page carries no effective date; not a commercial price.
Price. 160 USD per sample, triplicate, excluding the $300 setup (Scripps core rate card, fetched 2026-11-17, undated; as of 2026 (page undated))14
Reported, not yet replicated
Adaptyv Bio binding service (related measurement, not a plain SPR run) cloud-lab
Designed proteins are expressed and measured for binding on a biosensor; returns a binder call or kon, koff and KD. Its pages name both bio-layer interferometry and SPR, and the cross-lab page says single-cycle SPR. 161718
Caveats. Two live pages state different prices; minimum order is 20 or 24; terms promise only commercially reasonable efforts. Which instrument a given project uses is not public. Sequences only for designed proteins.
Price. 99 to 169 USD per protein or sequence (Vendor pages read 2026-11-17: "from $99", $149 with 2 replicates, $169 in the configurator; scope differs between pages.; as of 2026-11-17)1617
See: a record that is not published here
Reported, not yet replicated
Ginkgo Cloud Lab binder service (bio-layer interferometry, not SPR) cloud-lab
Cell-free expression of designed binders and bio-layer interferometry binding screen or kinetics, 5 to 8 business days. 19
Caveats. Bio-layer interferometry rather than SPR; 24-binder minimum; onboarding fee below 80 binders; target cost separate. Public price list only for this fixed protocol.
Price. 57 to 134 USD per binder, depending on test and batch size (Vendor price list, read 2026-11-17: $57 hit screen at 80 or more binders, $97 kinetics, $134 nanoDSF; $77 at the entry tier.; as of 2026-11-17)19
See: Ginkgo Bioworks
Reported, not yet replicated
Teach this
In one sentence
SPR measures how fast a molecule binds and lets go of a protein stuck to a gold chip, by watching the refractive index at the surface change. 1
Reported, not yet replicated
The five-minute version
Fix the protein to a gold surface, flow the compound over it, and watch the response rise as mass lands (association) and fall in buffer (dissociation). The curve is fitted for kon and koff, and KD is their ratio. Small molecules give tiny signals, so DMSO mismatches, sticky compounds and mass transport are the main ways it goes wrong. It proves direct binding to a purified protein, nothing about function or cells. In CACHE #1, SPR found 73 binders among 1,955 predicted compounds, at KD under 150 uM. 11015
Reported, not yet replicated
What a listener will ask
What does SPR give that an IC50 does not?
Direct binding and the rates kon and koff, so you can tell a long-lived complex from a short-lived one at the same KD.
Why is DMSO such a problem?
DMSO has a high refractive index, so a small mismatch between sample and running buffer gives a signal as large as the binding itself; a solvent-correction cycle is run for this.
Does a binder in SPR work in cells?
Not necessarily. SPR does not show function or cell entry, which is why cell target-engagement assays come next.
The common wrong idea
A good-looking sensorgram fit proves the kinetic constants are right.
It is deceptively easy to generate kinetic data: mass transport, rebinding and a damaged ligand can all produce a clean fit, so check at several ligand densities and flow rates.
The one citation to hand over: 1
Still open here
- Who first demonstrated SPR for biosensing, and when? The page the report relied on could not be opened, so names and year were removed.
- What is SPR's real throughput and per-compound cost? Public prices are academic core rates only; the Biacore 8K figure of more than 2,000 compounds a day is a vendor claim from a CRO page, Cytiva's own claim was not verified, and no independent throughput or per-compound price was found.
- How does SPR compare with bio-layer interferometry and grating-coupled interferometry? The only comparison found is by a competing vendor.
- CACHE Challenge #1: the full text was not opened, so the SPR protocol (concentrations, chip, counter-screens, second assays), the authors' conclusions and the ML-versus-docking comparison are unconfirmed.
- Whether array SPR (384 ligands per run) suits small molecules is not shown in the sources.
- The Biacore 8K product page and the exact solvent-correction setup (number of injections, 4-point setting) were not confirmed.
- No before-and-after (breakthrough watch), no capital entries and no per-campaign cost basis are written because the report does not support them.
- No source states a public data route for SPR sensorgrams or constants; only service routes are listed.
- Binding is not function: no source here says whether the CACHE binders are functional.
- Time scale of binding: the clock arithmetic (kon at the diffusion ceiling) is our calculation; in-cell free concentrations and competition at proteome scale have no opened source, so what a given koff means inside a cell is open.
Not yet written: Before after, Capital.
Linked from
- Co-immunoprecipitation (co-IP) and pull-down
- Crosslinking mass spectrometry (XL-MS)
- Dose-response assay
- Enzyme activity assay
- Isothermal titration calorimetry (ITC)
- Structural evidence of complexes (crystallography and cryo-EM)
- Target engagement in cells (CETSA, NanoBRET)
- Hit discovery
- Mechanism-confirmation assays
- Medicinal chemistry: potency, selectivity, ADMET
- Pathways and interaction partners
- Binding kinetics: on rate, off rate and residence time
- Ginkgo Bioworks
- Reaction Biology
- What would a CRO or cloud lab charge for protein-interaction assays?
- In a crowded cell, which partner wins when several proteins compete for one site?
- Has anyone measured kon and koff for interactions at proteome scale?
- Two price lists for the same binding experiment
- A hit series that binds and then fails in cells
- A target with no known partners
Sources
- Oxford Dunn School of Pathology. SPR guidelines (Sir William Dunn School of Pathology, University of Oxford). University of Oxford (2001). web, not peer reviewed; opened 2026-11-17 all records citing this
- Biacore (Wikipedia). Wikipedia (2026). web, not peer reviewed; opened 2026-11-17 all records citing this
- Biacore help: solvent correction procedure. Cytiva. vendor-page, not peer reviewed; opened 2026-11-17 all records citing this
- Duke Biomolecular Interaction Analysis core: services and rates. Duke Human Vaccine Institute. organisation-page, not peer reviewed; opened 2026-11-17 all records citing this
- Reaction Biology: surface plasmon resonance (SPR) service. Reaction Biology. vendor-page, not peer reviewed; opened 2026-11-17 all records citing this
- Schlosshauer, M., Baker, D.. Realistic protein-protein association rates from a simple diffusional model (arXiv q-bio/0401034). arXiv (2004). preprint, not peer reviewed; opened 2026-11-17 all records citing this
- Felder, S., et al.. SH2 domains exhibit high-affinity binding to tyrosine-phosphorylated peptides yet also exhibit rapid dissociation and exchange. Molecular and Cellular Biology (1993). paper; opened 2026-11-17 all records citing this
- Kastritis, P. L., et al.. A structure-based benchmark for protein-protein binding affinity. Protein Science (2011). paper; opened 2026-11-17 all records citing this
- Li et al.. CACHE Challenge #1: LRRK2 WDR domain hit finding (Li et al., 2024). Journal of Chemical Information and Modeling (abstract via Europe PMC) (2024). paper; opened 2026-11-17 all records citing this
- Tecan Journal: dramatically increasing throughput and efficiency of SPR analysis. Tecan. vendor-page, not peer reviewed; opened 2026-11-17 all records citing this
- Scripps Biophysics and Biochemistry Core: SPR. Scripps Research (2026). organisation-page, not peer reviewed; opened 2026-11-17 all records citing this
- Malvern Panalytical: BLI vs SPR vs GCI. Malvern Panalytical. vendor-page, not peer reviewed; opened 2026-11-17 all records citing this
- Scripps Biophysics and Biochemistry Core: services and prices. Scripps Research (2026). organisation-page, not peer reviewed; opened 2026-11-17 all records citing this
- NSF PAR record 10554378 for the CACHE Challenge #1 paper. NSF Public Access Repository (2024). web, not peer reviewed; opened 2026-11-17 all records citing this
- Adaptyv Bio home page (binding service price). Adaptyv Bio (2026). vendor-page, not peer reviewed; opened 2026-11-17 all records citing this
- Adaptyv Bio binding service page. Adaptyv Bio (2026). vendor-page, not peer reviewed; opened 2026-11-17 all records citing this
- Adaptyv Biosystems Foundry Terms and Conditions. Adaptyv Biosystems (2026). vendor-page, not peer reviewed; opened 2026-11-17 all records citing this
- Ginkgo Cloud Lab: binder services and prices. Ginkgo Bioworks (2026). vendor-page, not peer reviewed; 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 (6)
This record has been changed 6 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
- medium
- Emerging
- no
- Status
- partly written
- Last updated
- 2026-11-17
- Record
- content/assays/surface-plasmon-resonance.json