MACS Matchmaker

Interpreting Results

Scientific guidance for reading a finished measurement: which binding model fits the data, whether the fit can be trusted, what the features of a sensorgram mean, and what belongs in a published kinetic result. For the theory behind the models, see Biomolecular Interaction Analysis. For the software steps that produce the fits, see Data Analysis.

The order to read a result in

Each step below decides whether the next one is worth doing. Taken out of order — a model chosen before the controls are checked, a model swapped before the residuals are read — the fit still converges and the number it reports still looks like a measurement.

Each step decides whether the next is worth doing. The first two can end the analysis before any model is fitted — which is the point of running them first.

Look at the traces. Before any fit: does the baseline hold, does association rise and dissociation fall, do the molograms agree with each other. Artifacts are recognized here, not in the fitted parameters — see sensorgram artifacts.
Check the controls. The blank, the matrix blank and the non-binding mologram decide whether the response is the interaction or the assay. A result whose controls were not read is not yet a result.
Fit the simplest plausible model. Start at Langmuir 1:1 unless the chemistry rules it out.
Read the residuals and the parameter uncertainty. Random residuals and rates the data could actually resolve, not the fit score alone — see fit validation.
Change the model only on a specific residual pattern. A more complex model must be justified by the structure the simpler one left behind, and by a physical reason to expect it. More parameters always fit better; that is not evidence.
Confirm with replicates. A rate constant from one mologram, or from one preparation, is a single observation. What to report and how to qualify it is in reporting kinetic results.

The terms used below are defined in the quick reference on Biomolecular Interaction Analysis.

Binding Models

Five binding models are available in the kinetics evaluation dialogs. For the physics behind each one and interactive simulators, see Biomolecular Interaction Analysis → Common Binding Models. The table below maps the model names used in the software to their physical interpretation and the scenarios in which you should choose them.

Model	Physical meaning	When to use	Simulator
Langmuir 1:1	One analyte, one ligand, reversible.	Starting point. Clean single-exponential association and dissociation.	Open →
Langmuir 1:1 with Mass Transport	1:1 interaction where diffusion from bulk to surface limits the apparent on-rate. Adds k_t.	Residuals in association can't be cleaned up; curves change shape with flow rate.	Open →
Heterogeneous binding — two components	Sum of two independent Langmuir responses driven by the same analyte concentration, each with its own rates and capacity.	Use when one component cannot describe the data. The fit does not identify whether heterogeneity comes from the surface or the sample, and does not model separate analyte concentrations or competition.	Open →
Langmuir 1:1 — Partially Non-Dissociative	Irreversible binding — a fraction of bound analyte never dissociates.	Little or no dissociation observed. If mass accumulates across repeated injections, use the non-dissociative multi-relation model instead.	Open →
Langmuir 1:1 — Non-Dissociative Multi-Relation	Non-dissociative fraction fitted independently for each injection.	Residual-to-peak ratios change across injections, so one shared sticky fraction cannot explain the complete run.	Open →

Off-rate estimation uses the longest dissociation after the highest analyte concentration. The non-dissociative models include a retained baseline in this estimate. The heterogeneous models fit two binding components, but initialize both off-rates from one effective single-exponential estimate. Standard Langmuir and Langmuir with Mass Transport also use a single-exponential estimate. With Auto Initial Parameters enabled, this estimate initializes k_off and the detectability checks determine whether it can vary during the full fit. Manual initialization still assesses detectability for reporting, but uses your parameter values and bounds.

Fit Validation & Quality Criteria

Obtaining a fitted curve is not enough — the fit must be validated to ensure the kinetic parameters are meaningful. The following criteria help assess whether a model accurately describes the data.

Anatomy of a well-fitted 1:1 sensorgram: baseline, exponential association to equilibrium, and exponential decay during dissociation. Residuals should scatter randomly around zero.

From symptom to model

When residuals show structure, the shape of that structure points at the model. Start from the 1:1 Langmuir and work down this table; each row links to a simulator you can sweep to see whether the symptom reproduces.

What you see in the data	Likely mechanism or analysis	Available in	Simulator
Steady-state titration only — one response value per [A], no time resolution.	4PL equilibrium / Langmuir isotherm	Equilibrium Evaluation	Open →
Single exponential rise to plateau, single exponential decay back.	Langmuir 1:1	Determine Kinetics	Open →
Same [A] gives different curves at different flow rates — early association looks linear.	Langmuir 1:1 with mass transport	Determine Kinetics	Open →
Two clearly different relaxation timescales (fast then slow phase).	Heterogeneous ligand / 1:2	Determine Kinetics	Open →
A biphasic trace tracks sample purity or changes after size-exclusion purification.	Heterogeneous analyte / parallel reactions	Determine Kinetics	Open →
Baseline drifts up cycle-to-cycle — analyte never fully washes off.	Partially non-dissociative 1:1	Determine Kinetics	Open →
The residual-to-peak fraction changes from one injection to the next.	Injection-specific non-dissociative 1:1	Determine Kinetics	Open →
Identical injections give smaller and smaller plateaus over a long run.	Decaying-surface 1:1	Simulator only — no production fitter	Open →

Residual Analysis

Residuals are the differences between the measured data and the fitted curve at each time point. A good fit produces random residuals — scattered evenly above and below zero with no discernible pattern. Systematic residuals (curved patterns, consistently positive during association and negative during dissociation) mean the model does not adequately describe the data. Common causes:

Wrong binding model (e.g. 1:1 Langmuir applied to a heterogeneous interaction).
Mass transport limitation not accounted for.
Baseline drift or injection artifacts not removed during data cleanup.

Read the residuals, not the overlay. A fitted curve can sit convincingly on the data while the residuals still carry the shape of the interaction. Scatter with no structure means the model has accounted for everything it can; an arc that is positive through association and negative through dissociation means it has not.

Parameter Sanity Checks

After fitting, verify that the returned parameters are physically plausible:

Parameter	Typical range	Out-of-range meaning
k_on	Compare with the expected interaction class	Very high values often indicate mass transport limitation; very low values suggest a non-1:1 interaction.
k_off	Resolvable within the measured dissociation window	Very small values should be cross-checked against the 5% decay rule — if dissociation was insufficient, the fit is unreliable.
R_max	Close to observed maximum response	Fitted R_max much larger than the observed peak (e.g. 10×) usually means the model is inappropriate or the data never reached saturation.
K_D	Agrees with steady-state isotherm K_D (if measured)	Material disagreement between kinetic and steady-state K_D calls for checking equilibration, mass transport, concentration accuracy, and model choice.

Chi² Interpretation

The reduced chi² value is the squared residual per degree of freedom, in the squared units of the response. Compare √chi² with the baseline noise measured in the same run and response quantity. A residual scale well above that observed noise indicates that the model or data should be re-examined.

When to Switch Models

If the Langmuir 1:1 model produces systematic residuals, match the symptom you see to the right model using the decision table at BIA → Choosing a Model. Start simple; only move to a more complex model when residuals or the biology of the interaction justify it.

Sensorgram Interpretation & Artifacts

Understanding common sensorgram features and artifacts helps distinguish real binding events from instrumental or experimental effects.

flat baseline, smooth rise, clean decay

ligand dissociation, or matrix binding to ridge or groove

normal at the injection boundaries

ligand loss, depletion, or a degrading sample

MD climbs while CMD barely moves — the mass is not on the ridges

One defect at a time. Every panel is the same 1:1 interaction; the dashed line is the trace it should have produced. The shaded band is the injection. Spikes at the two injection boundaries are normal and are trimmed during cleanup; a baseline that climbs through the whole run, a response that falls while sample is still flowing, or a signal that appears in MD but not CMD are not artifacts of the instrument but statements about the surface or the sample.

Common problem signs and what to do about each:

Symptom	Likely cause	Possible solution
Baseline rises or falls before or between injections	Ligand slowly dissociating from the surface, or a matrix component binding to the ridges or grooves (also temperature drift or buffer mismatch)	Use a more stable or covalent immobilization if the ligand is dissociating, and improve affinity-matched backfilling if a matrix component is accumulating; equilibrate the instrument and match buffers to rule out drift.
Sharp spikes at the start and end of the injection	Normal refractive-index transient at the injection boundaries, or a drastic change in flow rate	Boundary spikes are expected and trimmed automatically during cleanup; avoid large step changes in flow rate between phases.
Response falls while the sample is still injecting	Sample dilution from too small a pickup volume, or ligand loss or a degrading sample	Increase the pickup volume (or use Inject and Incubate), and check the ligand stability and the sample integrity.
Signal appears in the refractometric channel (MD) while the diffractometric channel (CMD) barely moves	Mass adsorbing off the ridges — non-specific binding rather than a specific interaction	Improve affinity-matched backfilling, and confirm with control molograms and buffer-only blank injections.
Association bends away from a 1:1 (Langmuirian) shape and the fit leaves structured residuals	Diffusion / mass-transport limitation — analyte binds faster than it reaches the surface	Use a faster association flow rate; if it persists, lower the ligand density or fit a transport-aware model.

Baseline Drift

A gradually rising or falling baseline before or between injections can be caused by:

Ligand dissociation: A ligand slowly falling off its immobilization site makes the baseline decline over the run; use a more stable or covalent attachment.
Matrix binding: A sample-matrix component binding to the ridges or grooves makes the baseline climb; improve affinity-matched backfilling and use control molograms to confirm.
Incomplete backfilling: If the grooves are not adequately passivated, non-specific binding can accumulate coherently. Ensure the backfilling level matches the frontfilling level (see Backfilling and NSB suppression).
Temperature drift: Allow the instrument to equilibrate for at least 15 minutes before starting measurements.
Buffer mismatch: Focal molography is more tolerant of buffer mismatch than SPR, but differences between running buffer and sample buffer can still cause bulk transients or gradual signal changes. Minimise mismatch whenever you need quantitative kinetics or when bulk effects are visible.

Injection Artifacts (Spikes)

Some overshoot at the start and end of sample injection is normal in FM sensorgrams. Spikes are visible in the refractometric channel and may propagate into the diffractometric channel as brief transients. During data cleanup, remove them with the "Remove spikes" action or by trimming injection boundaries.

Declining Response During Injection

If the signal rises initially but then decreases while the sample is still flowing, this typically indicates sample dilution due to insufficient pickup volume. The trailing edge of the sample plug becomes diluted through diffusion in the tubing. Increase the pickup volume or refer to the Inject and Incubate guidance in the Experiment Design documentation.

Distinguishing Specific from Non-Specific Binding

FM's diffractometric channel inherently rejects random (incoherent) non-specific binding. However, coherent NSB can still occur if backfilling is not affinity-matched. To distinguish specific from non-specific signals:

Compare the diffractometric (coherent) and refractometric channels. A large refractometric signal with minimal diffractometric response indicates predominantly non-specific binding or a bulk effect.
Use control molograms (backfilled without active ligand) as on-chip references. Binding on control molograms indicates non-specific interaction.
Perform blank injections with buffer only to confirm no response is observed.

What Good Data Looks Like

Before fitting kinetics, inspect the sensorgram visually. The table below summarises what to look for — and what to do when you see the problem sign instead.

Feature	Good sign	Problem sign → likely cause / next action
Baseline before injection	Flat with only small random fluctuations.	Sloping baseline → temperature drift, incomplete backfilling, or residual contamination.
Association phase	Smooth, consistent increase across active molograms during sample injection.	Signal rises then falls during the same injection → insufficient pickup volume, sample plug diluted.
Dissociation phase	Measurable decay, or stably high in a way that matches the expected interaction type.	No decay on a reversible interaction → regeneration may be needed; non-1:1 behavior possible.
CMD vs refractometric channel	Coherent (CMD) signal tracks specific binding; refractometric channel clean.	Strong refractometric change with little CMD response → mostly bulk or NSB, not coherent binding.
Regeneration	Baseline returns close to pre-binding level before next cycle.	Poor return → incomplete analyte removal or surface fouling; consider an escalated regeneration condition.
Mologram consistency	Individual molograms follow the same trend; consensus trace represents the population.	Large mologram-to-mologram disagreement → surface inhomogeneity, bubbles, edge effects, or outlier molograms to exclude.

Reporting Kinetic Results

When publishing or sharing kinetic data from the MACS Matchmaker, include sufficient detail for others to evaluate the quality and reproducibility of the measurements. The checklist below covers the minimum information to report.

Kinetic Parameters

For complete kinetics: k_on (M⁻¹s⁻¹), k_off (s⁻¹), and K_D (from k_off/k_on).
For association-rate-only results: the fitted k_on and its uncertainty interval when available, k_off as below the reported detection limit, and K_D as unavailable. Include the observed dissociation decay and the required detection criterion.
Fitted R_max — dimensionless when normalized to immobilization, which is the default, otherwise in pg/mm² — and the fit-quality evidence.
If equilibrium analysis was performed: K_D from the steady-state isotherm and whether it agrees with the kinetic K_D.

Experimental Conditions

Binding model used (e.g. Langmuir 1:1, Heterogeneous binding) and justification for the choice.
Chip type (DDI or click chemistry) and surface chemistry.
Immobilization method, ligand identity, and capture level (pg/mm²).
Analyte identity, concentrations used, and number of replicates.
Kinetic format (SCK or MCK) and flow rates for association and dissociation.
Running buffer composition and temperature.

Data Quality Evidence

Sensorgram overlay showing measured data and fitted curves for all concentrations.
Residual plot demonstrating random (not systematic) deviations.
If multiple replicates were performed, report the mean and standard deviation of kinetic parameters across independent experiments — not the standard error from a single fit.