Common Issues in Kinetic Data
Even the best designed experiments can suffer from artifacts. Identifying them early saves you from fitting wrong models to bad data. Here is how to spot the most common culprits.
Interactive Diagnostics
Select a scenario to see how it affects the sensorgram.
Ideal Data
Perfect exponential curves fitting the 1:1 model.
Simulated parameters:
In Conclusion: What to look for before fitting?
Verify your data against this quality checklist. Fitting models to low-quality curves will give wrong answers and waste time.
✓ High Quality Checklist
- •Flat Baseline: Signal drift is close to zero when buffer is flowing.
- •Low Buffer Jump: RI shift upon injection start/stop is very low.
- •Exponential Curvature: Association shows clear curvature (not linear), indicating binding is reaction-limited, not transport-limited (free of mass transport).
- •Sufficient Curvature: Data has enough "shape" to define constants.
- •Replicates: At least one concentration is repeated to check stability.
- •Dynamic Range: Concentrations span 0.1 – 10x expected KD.
- •Well Spaced: Curves are clearly separated in the sensorgram.
- •Steady State: At least one curve reaches equilibrium (if possible).
- •Dose-Response: Higher concentrations give higher responses following the expected Langmuir isotherm (hyperbolic, not linear).
- •Long Dissociation: Decay is long enough for reliable koff.
- •Randomized: Injections are randomized to avoid systematic errors.
🛠️ Steps for Optimization
If data is low quality, optimize conditions before fitting.
- 1.Check ligand and analyte for purity and uniformity (SEC).
- 2.Lower Ligand Density on the sensor chip (reduces MTL).
- 3.Increase flow rate to check/reduce mass transport (SPR only — for BLI, increase shake/agitation speed, typically to 1000 rpm; there is no flow during association/dissociation).
- 4.Reverse ligand and analyte on the sensor chip.
- 5.Change immobilization chemistry (e.g., thiol or capture).
- 6.Use site-directed coupling or capture the ligand.
- 7.Optimize buffer: adjust salts or detergents (Tween-20).
📉 Baseline Drift
What it looks like: The response signal steadily increases or decreases even when no analyte is being injected (e.g., during buffer flow).
Common Causes:
- Dissociation of Ligand: If the ligand is not covalently attached (e.g., captured via His-tag), it may slowly leak off the surface.
- Temperature Fluctuations: SPR is extremely temperature sensitive. Ensure the instrument is equilibrated.
- Incomplete Regeneration: Accumulation of analyte on the surface over time.
- Buffer Mismatch / Bulk-RI Re-equilibration: After a buffer change, the system can take many minutes to settle as the running buffer equilibrates throughout the fluidics and matrix.
- Non-specific Binding on the Reference Surface: NSB accumulating on the reference channel produces drift that is notcommon-mode and cannot be subtracted away.
How to fix: Use "Double Referencing". Subtract the signal from a reference channel (no ligand) AND subtract the signal from "blank" injections (buffer only). This removes linear, common-mode drift — but non-linear drift on the active channel only (e.g., ligand leak) survives double referencing and must be addressed at the source (covalent coupling, longer equilibration, better blocking/Tween-20 to suppress NSB).
Bulk shifts & solvent correction
When the analyte buffer differs in refractive index from the running buffer, you get a square “bulk shift” at injection start/stop on top of the binding signal. Common causes and fixes:
- DMSO / cosolvent mismatch (dominant artifact for fragment and small-molecule work): even a 0.1% DMSO difference between running buffer and analyte buffer produces a bulk shift far larger than the binding response. Run a solvent correction curve (a DMSO calibration series across the expected range) and apply it before fitting.
- Excluded-volume mismatch between reference and active surfaces when ligand densities differ greatly — the reference does not see the same bulk-RI environment, so subtraction is imperfect. Match reference surface chemistry/blocking to the active surface as closely as possible.
- Injection-boundary spikes from misaligned reference subtraction (timing offset between channels). These manifest as sharp transients at t = 0 and at the end of injection; check channel alignment in the instrument software.
Baseline Drift Simulator
Explore how baseline drift affects your kinetic data and how to correct it
How Drift Distorts Sensorgrams
Solid lines = with baseline drift. Dashed = ideal (no drift). Drag the slider to see how drift accumulates over the measurement time.
🚚 Mass Transport Limitation (MTL)
What it looks like: The association curve is linear (straight line) rather than exponential curvature. The observed binding rate depends on the flow rate.
Why does this happen?
In SPR and BLI, the analyte must physically travel from the bulk solution to the sensor surface before it can bind. When the binding reaction is very fast (high ka) or the surface has many binding sites (high Rmax), the analyte near the surface gets consumed faster than it can be replenished. This creates a depletion zone — a region of low analyte concentration directly above the surface.
The result: instead of measuring the true binding rate, you measure the rate of analyte delivery to the surface. The association phase becomes linear (limited by diffusion) rather than exponential (limited by the reaction).
The Damköhler Number
The severity of mass transport limitation is captured by a single dimensionless number — the Damköhler number (Da):
Da ≈ ka × [B]surf / kt
where [B]surf is the surface concentration of binding sites (mol/m², for which Rmax in RU is a convenient proxy) and kt is the mass transport coefficient in m/s (depends on flow rate and flow cell geometry). This is a scaling expression — useful for ranking risk, not as a numerical calculator.
Note: kt scales only weakly with flow (Lévêque: kt ∝ Q1/3), so a 3× increase in flow rate buys only ~1.4× in kt.
- Da < 0.1 — No significant MTL. Your kinetics are reliable.
- 0.1 < Da < 1 — Moderate MTL. ka will be underestimated.
- Da > 1 — Severe MTL. The observed rate is dominated by diffusion, not binding.
⚠️ Who is most at risk?
Fast binders (ka > 10⁵ M⁻¹s⁻¹) on surfaces with high binding-site density are most susceptible. The relevant quantity is the surface site density (roughly > 1 fmol/mm²), not Rmax in RU per se — for a small analyte, 100 RU represents many more sites per area than it does for a 150 kDa antibody. As a rule of thumb: Rmax > 100 RU with a small analyte, or any surface where the analyte is amine-coupled at high density, warrants suspicion. This includes many antibody–antigen interactions.
How large is the error?
This is the part that often surprises people. Mass transport limitation doesn't just add noise — it systematically distorts your reported kinetics:
- ka is always underestimated — because the apparent association rate is capped by the rate of analyte delivery. With severe MTL, your reported kacan be 10–100× lower than the true value.
- kd is typically underestimated — rebinding (dissociated analyte re-binding to a neighboring site before escaping the depletion zone) makes the apparent off-rate appear slower than the true kd. This partially offsets the ka error and can make the data look deceptively self-consistent.
- KD appears weaker — since ka is underestimated more than kd, the calculated affinity (KD = kd/ka) shifts toward weaker binding. You might report 10 nM when the true affinity is 1 nM.
🚨 The hidden danger
MTL-affected data can still look like it fits a 1:1 model reasonably well, with decent χ² values. The fit “works” — but the parameters are wrong. This is why the flow rate test is essential: you can't diagnose MTL from the goodness of fit alone.
The Golden Rule check:
If you increase the flow rate and the binding curve changes (gets faster), you have Mass Transport Limitation. If the curves overlay perfectly, you are in the clear.
How to fix it
- Lower the ligand density — This is the most effective fix. Aim for Rmax < 50 RU for antibodies. Use the simulator below to find your target.
- Increase the flow rate — Higher flow rate increases kt, reducing Da. Go from 30 to 50–100 µL/min if your instrument allows it.
- Fit with a mass transport model — Adds kt as a fitting parameter. This can recover the true ka, but only if the MTL is moderate. With severe MTL, even the MT model struggles.
- Switch the assay orientation — Immobilize the larger molecule and flow the smaller one. Smaller analytes diffuse faster (higher kt).
Mass Transport Simulator
Explore how mass transport limitation affects your kinetic data
Ideal vs. Mass Transport Limited
Solid lines = with mass transport. Dashed = ideal 1:1. Notice how MTL makes the association phase linear and slows apparent binding.
🧩 Heterogeneity
What it looks like: The data cannot be fitted well by a 1:1 model. Usually, the dissociation phase has a "fast" initial drop followed by a "slow" tail.
This suggests there are multiple populations of binding sites, each with different kinetics.
Sources of Heterogeneity:
- Chemical Conjugation: Amine coupling randomly orients proteins, hiding or exposing binding sites differently.
- Aggregates: The analyte sample contains dimers or aggregates. In antibody work this is the most common cause — and it is often introduced during dilution into running buffer, not in the stock. Run SEC immediately before the experiment, not just on the stock material.
- Polyclonal Antibodies: Naturally contain a mix of affinities.
How to fix: Try oriented coupling methods (e.g., Biotin-Streptavidin, His-capture). Ensure analyte quality (SEC). If unavoidable, use a "Heterogeneous Ligand" model to fit the data.
Heterogeneity Simulator
Explore how multiple binding populations affect your sensorgram
1:1 vs. Heterogeneous Binding
Two independent populations with different kinetics. Solid = combined signal. Dashed = individual populations. Watch for the fast drop + slow tail in dissociation — the hallmark of heterogeneity.
🔄 Rebinding & Avidity
Two related artifacts that make your KD appear tighter than reality. Both artificially slow the apparent kd (off-rate), leading to overestimated affinity.
What is Rebinding?
An analyte molecule dissociates from one ligand but re-binds to a neighboring ligand before diffusing away. This artificially slows the apparent kd, making KD appear tighter than reality. It is a surface density artifact — the higher the ligand density, the worse the rebinding.
What is Avidity?
When a multivalent analyte (e.g., IgG with 2 binding arms) binds multiple ligands simultaneously. The effective off-rate is much slower because both arms must release simultaneously. The measured KD reflects the "apparent" affinity, not the intrinsic monovalent affinity.
How to Diagnose:
- Compare kinetics at 2–3 different ligand densities (e.g., 20 RU, 50 RU, 100 RU).
- If kd changes with density → rebinding is present.
- If kd is the same across densities → true kinetics.
- Concentration-dependent kd is another red flag (off-rate should be independent of analyte concentration).
How to Fix:
- Lower ligand density — same canonical target as for MTL: Rmax < 50 RU for antibodies on SPR (≈ 0.3–0.5 nm for BLI). Spacing the ligand far apart is the primary lever against bivalent-analyte avidity.
- Increase flow rate (SPR only) — sweeps dissociated analyte away faster, reducing rebinding.
- Use monovalent analyte — Fab fragments instead of full IgG to eliminate avidity.
- If the ligand is itself bivalent (e.g., capturing an IgG as ligand), orient it so only one binding site is presented. Note: when avidity comes from a bivalent analyte binding two surface ligands, the lever is ligand spacing (low density, see above) — orienting the ligand does not help because the analyte still has two arms.
Most affected: IgG antibodies, any bivalent or multivalent molecule, and high-density surfaces. If you are measuring antibody kinetics, always check for avidity by comparing Fab vs. IgG or by varying ligand density.
Rebinding & Avidity Simulator
Explore how rebinding and avidity distort your kinetic measurements
Rebinding Effect
Higher ligand density = more neighboring sites for rebinding. The apparent kd slows as density increases, making KD appear tighter than it really is. Purple dashed = ideal (no rebinding).
♻️ Regeneration Failure
What it looks like: Data quality degrades across cycles. The fix depends entirely on which baseline metric is drifting — these have opposite causes and opposite fixes.
Declining Rmax across cycles
The initial response at a fixed analyte concentration falls cycle-over-cycle. This is ligand damage — the regeneration condition is too harsh and is denaturing or stripping active ligand. Fix: soften the regeneration (shorter contact time, lower pH/concentration), or switch to a capture format (His, anti-Fc, biotin/SA) so you re-load fresh ligand each cycle instead of regenerating it.
Rising baseline across cycles
The baseline climbs cycle-over-cycle while Rmax holds steady. This is carry-over — regeneration is too weak and analyte is accumulating on the surface. Apparent kd looks slower than it really is, and the signature can mimic heterogeneity. Fix: scout a stronger regeneration solution.
Regeneration scouting protocol
When designing a regeneration, run a scouting matrix before committing to a method:
- Test low pH (glycine pH 2.5–1.5), high pH (NaOH 10–50 mM), high salt (NaCl / MgCl2 1–4 M), and detergent (0.05% SDS) — alone and in combinations.
- Run 3–5 consecutive cycles with each candidate and track both the baseline trajectory and the Rmax trajectory. Accept only conditions where both are stable.
- Watch for cycle-to-cycle kd drift in the fitted parameters — a slowly rising apparent off-rate is often the earliest sign of carry-over masquerading as good data.
Alternatives to regeneration
If no regeneration solution preserves both baseline and Rmax, switch formats. Capture surfaces (His-tag/NTA, anti-Fc, biotin/SA) let you discard the ligand each cycle and re-load fresh — bypassing the regeneration problem entirely at the cost of a re-loading step. For BLI, this is the default workflow. For SPR, it is the standard escape hatch for fragile ligands.
Have SPR or BLI data?
Upload your raw files and get an automated kinetic analysis in minutes. We support Biacore, Octet, and other major formats.
Upload & Analyze