Workflows

The project home groups available evaluation goals into Kinetics and Affinity, Concentration Determination, and Limit of Detection. Open Kinetics and Affinity to start a guided study setup, name and tag the study, choose between Binding Kinetics, Equilibrium Analysis, and DEL Evaluation, and then continue through the setup directly on the page. Open Limit of Detection to enter the evaluation name, tags, and description, select eligible raw data, processed data, or an existing evaluation, configure the response and fit parameters, and review the key settings before starting the evaluation. Binding Kinetics studies ask whether to run Determine Kinetics, SCK, or MCK before data selection. Every guided evaluation accepts the compatible source types shown in its Data step. While compatibility is checked, each row shows a spinner in a fixed-size selection control; unsuitable rows remain visible but disabled, and their selection controls show the validator message that explains which requirement the data does not meet. Raw data can also be added to the project directly from the Data step. Data-selection tables reserve their widest column for the source name and omit the description column so long names remain identifiable. The standard stepper shows Goal, Data, Parameters, and Review. Equilibrium Analysis shows Goal, Data, Cleanup, Parameters, and Review. Determine Kinetics shows Goal, Data, Cleanup, Parameters, Fitting, and Review so cleanup always comes before the fitting setup. Step numbers stay aligned when labels wrap on a narrow screen. The footer groups Cancel, Back, Next, Run, or Save actions on the left. Once anything has been entered, Cancel asks for confirmation before it leaves the workflow, because leaving discards the setup. The same confirmation appears when the workflow is left in any other way, such as the browser back button or a reload, while an untouched workflow is left without a question. Review summarizes the key setup values, including the names of cleanup actions inherited from selected data and applied by the workflow. It displays the first three action names and an ellipsis when more are present; select the summary to show or hide the complete list. The final Run action starts and saves every evaluation. Concentration Determination opens a guided workflow for naming the study, selecting calibration data and unknown data together, cleaning each selected input, configuring prediction parameters, previewing the generated plots, and optionally saving the evaluation. Its steps are Details, Data, Cleanup, Parameters, Fitting, and Review. A calibration measurement with an Unknown Concentration phase selects itself as the unknown automatically.

Re-evaluating an evaluation opens the same guided setup it was built in, with its data, parameters, description and tags filled in and its name carrying the next version: re-evaluating Sample proposes Sample (Version 2), and re-evaluating that proposes Sample (Version 3). A setup opened this way, or from a Use for … Evaluation button on processed data, starts past the steps its link has already answered: at Cleanup for Determine Kinetics, at Fitting for Concentration Determination, and at Parameters for every other evaluation. The earlier steps stay open in the stepper. Concentration Determination fills its calibration and its unknown role separately, each from the raw data that role ran on. Determine Kinetics also restores the fits you accepted and rejected by hand, together with the fit rejection strictness, so its fitting step opens where you left it. What the strictness decided is not carried over but worked out again from the new fits, so a fit rejected automatically comes back as an automatic rejection you can still overrule, and one that no longer converges is marked invalid. Selecting an evaluation as the data source in a Data step contributes its data only, together with the data selection and cleanup actions stored with it; the parameters stay at the defaults for that data. The Evaluations list in a Data step offers only evaluations of the kind being set up, so the data it hands over is data that kind of evaluation has already run on.

Common settings stay visible in the main parameter cards. Less frequently changed settings are collected in Advanced Parametersand collapsed initially. Equilibrium Analysis configures interactive cleanup actions and automatic cleanup in its Cleanup step. SCK and MCK keep Auto Data Cleanup and the actions it controls together in a dedicated Data Cleanup card in Parameters. Parameter labels include the scientific unit or state that a value is a fraction or dimensionless. Evaluation plots are generated automatically, so the guided workflows do not require a separate plot checkbox.

Binding Kinetics modes

• Determine Kinetics: clean and fit one dataset interactively to estimate kinetic rate constants and affinity. Use Cleanup to review cleanup actions and Fitting to inspect and accept or reject fits. The info button at the right of Model Summary opens the interactive simulator for the selected binding model.
• SCK (Single-Cycle Kinetics): inject increasing concentrations in one cycle without regeneration between injections. This uses less sample and finishes faster.
• MCK (Multi-Cycle Kinetics): measure one concentration per cycle and regenerate between cycles for independent binding cycles.

The Kinetics and Affinity wizard includes an Evaluation Assistant when the shared assistant service is enabled. It starts with the current goal, step, and data selection so you can ask about model choice, data requirements, cleanup, fit acceptance, or review. Its answers use the same grounded product and science documentation as the Documentation Assistant.

Use the Project Overview metrics to jump from the home page to Raw Data, Process Data, Responses, Evaluations, or Report Points. To assign measurements to the project, open Raw Data and use the add (+) action there. Data steps that accept measurements, including Kinetics and Affinity and Limit of Detection, also offer an Add Raw Data to Project action.

Raw Data plots show all signal molograms by default for multi-channel flow chambers and multiplexed measurements. A 1-plex Sirius 8×8 measurement made with the single flow chamber starts with the inner signal molograms selected; the mologram picker can change either selection.

Configure Report Points

The Configure Report Points dialog previews the report points for a measurement before you rely on them. Open it from a plot on the Perform page, or from the More actions menu of any entry in a project's Processed Datatable. When opened from processed data, the entry's clean-up actions are applied first, so the plot (with report-point markers), the Report Points Table, and the export all reflect the processed (cleaned) data. In that mode the phase comes from the processed-data selection and the plot uses the marker-supported consensus view; holograms, injection-type settings, and the back-fill threshold remain editable.

CSV export

The dialog-level Export CSV button exports all sensor valuesalongside the consensus summary: the file contains one row per report point and sensor, each with its own mean, median, standard deviation, maximum, minimum, and slope, plus the report point's consensus (over-all-sensors) median, mean, and slope repeated on every row. Every row also carries the measurement name, sensor name, ligand name, the report-point timestamp, and the selectedtime window (s). The per-sensor mean is the arithmetic mean of that sensor's values inside this window; the consensus mean is then the mean of the per-sensor means. This makes it possible to trace every exported mean back to the interval used for its calculation while the consensus median remains available.

The Report Points Table also includes a Show Details action for each row. It opens a per-report-point view with a sensor heatmap, a sensor-value boxplot, and a CSV export containing only that report point's per-sensor rows.

Save to project

Automatically detected report points can be saved so they persist and appear on the Report Points page. Click the save icon in the Report Points Table header to reveal a checkbox on each savable row, select one or more report points, and click Save to Project. When the dialog is opened inside a project the report points are saved to that project directly; otherwise you pick the target project from a dropdown. Manually created report points and automatically detected points that were saved previously are already stored and cannot be re-saved. If part of a multi-point save fails, only the failed report points remain selected for retry.

Report Point Calculations

For each report point, the system evaluates the selected report-point time window (for example the configured 5-second window near an injection end). It first computes a consensus trace (also called median trace) by taking the median value across all sensors at each time point — a robust central estimate that is less affected by outliers. All report-point values are then derived from the consensus trace and the individual sensor traces:

• Value: the consensus trace value at the report-point timestamp.
• Slope: the mean of slopes computed across all individual sensor traces within the report-point time window (linear regression per sensor; arithmetic mean across sensors).
• Delta:the difference between this report point's value and the previous report point's value in the sequence.

Automatic Report Points

When viewing a measurement with injections, report points can be generated automatically based on injection type:

Response Plot Units

Evaluation plots label and scale the response axis from the response quantity. Normalized responses are dimensionless and are shown as normalized (a.u.). Coherent mass density is shown in pg/mm² using one conversion from its base unit to the display unit. Measured points and fitted curves therefore use the same scale.

Evaluation Result Findings

Completed SCK, MCK, Determine Kinetics, DEL, and Concentration Determination evaluations display a compact, result-specific findings panel automatically. A stateless findings service reads the evidence saved by the evaluation. It does not invoke an LLM, rerun the evaluation, or parse a generated report.

SCK, MCK, Determine Kinetics, and DEL use the same compact report: accepted and rejected counts appear below the findings title, outcome rows state the conclusions, and bounded exceptions or recommended actions are available under Technical details. The saved observations include:

• Result summary: how many fits or ligands were accepted and rejected. Only accepted results are used in the conclusions.
• Fit quality:a short ligand-level conclusion naming the model and confirming that accepted results passed the evaluation's fit-quality gate. If k_off was estimated from a dissociation segment and held fixed in the global fit, the finding still reports a successful fit while qualifying k_off and K_D. Clean SCK/MCK design reports do not repeat this row.
• Main binding results: for multiplex chips, which ligands have similar on/off-rate behavior. A ligand with only one accepted sensor or an unresolved off-rate is left unclassified.

For an SCK or MCK evaluation using the strict Langmuir 1:1 model, the evaluation checks every accepted fit against the concentration series, programmed association and wash durations and final dissociation recorded for that result, plus its measured sampling interval. Autosampler pickup gaps may extend the plotted trace intervals but do not change the programmed durations checked here. When every accepted fit forms an exact design request, the findings cover equilibrium suitability, 1:1 fit design, association timing, dissociation timing, and any fitted rate outside the measurable range. This deterministic guidance replaces the generic concentration and dissociation suggestions instead of duplicating them. If any accepted fit is not exactly representable, the complete design section is omitted and the generic concentration and plateau cautions are used.

Concentration Determination uses the same deterministic findings panel with concentration-specific evidence. Its summary reports how many unknown/dilution prediction groups were attempted and contributed to the combined estimate. Estimates inside the final fitted calibration interval use interpolation. A response outside the fitted values at those retained-standard boundaries has no predicted concentration and zero effective weight; it is never extrapolated or endpoint-clamped. Each sensor fit is plotted only over its retained-standard interval. When relevant, the range finding reports below-range, above-range, and unavailable sensor counts with bounded per-group detail, including each affected sensor's measured response and status. Fitting contains the calibration curve, boxplot, and selectable per-injection tables; Review contains the study and numeric result summaries plus warnings. The completed result uses this findings panel as its single diagnostic summary instead of repeating those workflow warning alerts. For logistic models, a separate calibration finding warns when any assessable curve reaches less than 90% of its fitted high-concentration plateau or a curve cannot be assessed; it reports the affected curve count and, when a curve is assessable, median coverage, threshold, and the highest retained standard. This plateau assessment does not apply to the non-saturating cubic model.

A separate dilution-response-compression check examines the three highest sample fractions only when they span at least 4× in dilution factor. It normalizes each sensor's responses to the fitted response interval between the lowest and highest retained calibration standards. A sensor is compressed when all three responses stay at or above the 75% calibrated response position and span no more than 15% of that interval; the warning appears when at least four sensors are assessable and at least half are compressed. This is an advisory warning for possible response saturation or dilution-series inconsistency and does not change exclusions or weights.

An older result that does not contain the saved evidence shows an explanation that the evaluation must be re-run or re-evaluated. Missing saved evidence does not mean that every ligand or prediction was rejected.

For supported kinetic and DEL evaluations, Ask appears with an AI sparkle icon in the findings header when the shared LLM API is configured and opens a right-side conversation. The findings remain visible without an LLM; only Ask is omitted. Concentration Determination displays deterministic findings without offering Ask. The assistant always receives the same bounded findings shown in the panel and a catalog of the names in this result. When a question needs more, it requests fit details, rejection evidence, or derived responses, and those lookups run on the Hub against this evaluation only. Every returned fit row is marked accepted or rejected; rejected rows explain exclusions only. The tools are read-only and cannot change stored data. With exact rate, timing, concentration, mode, and sampling inputs, the assistant can also run the same deterministic SCK/MCK design check as the phase editor. Missing inputs are requested rather than replaced with defaults. That what-if uses the same finding structure, carries no plot arrays, does not replace the saved evaluation evidence, and cannot apply suggested experiment changes. Artifact file contents are not treated as inspected.

Each conversation belongs to the person who started it and to one evaluation. A question and its answer are saved together once the answer finishes, so an interrupted answer leaves the conversation unchanged.

Exact values, plots, residuals, and sensor tables remain in the evaluation result. Findings use count-first outcomes and keep bounded exception evidence under Technical details; they contain no plot data. The assistant receives the same compact records. Multiplex groupings describe this measurement only; they do not establish biological classes.

DEL and Concentration Determination create this evidence from the same processed result used to build their saved artifacts. If a future conclusion needs a value that is not persisted, that deterministic calculation is added to the evaluation first. The findings service only selects, compares, and explains stored evidence.

SCK (Single Cycle Kinetics) Evaluation

Add your data to a project and create a processed-data element with the SCK phase. Click USE FOR SCK EVALUATION to open the guided setup on its Parameters step, with that processed data already selected and named after it, choose the model for fitting, and run. Results appear in the evaluation view. If you have cleaned up the data manually, do not use the auto-cleanup button — your changes will be overwritten. Every optimizer-converged SCK fit remains in the saved results and standard fit artifacts. SCK does not apply a final fit-rejection strictness filter. The result view adds a tagged Rejected fits artifact only when fitting does not produce a usable result for a mologram; it can be shown or hidden with Filter by tags. Slow-dissociation results remain successful and add a warning that recommends a longer dissociation phase. For existing data, correct only visible drift or baseline offsets and exclude injections with spikes or other artifacts; cleanup cannot compensate for a dissociation phase that is too short. Auto Data Cleanup and its available actions are kept together in an outlined group, while initialization and drift correction are in the collapsed Advanced Parameters section. Manual initial-parameter fields identify k_on in M⁻¹ s⁻¹, k_off in s⁻¹, normalized Rmax as dimensionless, and the normalized mass-transport constant k_t in M⁻¹ s⁻¹.

For the physics of each model, see the Binding Models table below or use the info button at the right of Model Summary to open the simulator for the selected model.

Slow dissociation warnings

SCK assigns each sensor fit one of three classifications:

• Complete kinetics — both k_on and k_off are quantifiable. The result includes k_on, k_off, and K_D.
• Association rate only — k_offis below the dissociation phase's detection limit. The fit remains successful: its fitted curves and k_on remain available, with the fitted uncertainty interval when it can be calculated. The result reports k_off as < detection limit and K_D as unavailable.
• Unusable — fitting did not produce a successful result for the mologram. It is excluded from saved kinetic results and listed in Rejected fits with the failure reason.

The k_off detection limit uses a fixed 5% criterion. The longest fitted dissociation must support at least a 5% signal decrease for k_off to be quantifiable:

k_off detection limit = −ln(0.95) / dissociation duration

A converged SCK fit is not passed through a final fit-quality or strictness rejection filter. A low k_off is therefore warning-only, and no additional option is required. The result page keeps the standard fit artifacts and adds a k_off warning with the observed decay and the recommended longer dissociation phase.

MCK (Multi Cycle Kinetics) Evaluation

Similar to SCK but intermediate regenerations are performed before each cycle. Choose the model to fit and run.

Equilibrium Evaluation

Equilibrium evaluation extracts a steady-state dissociation constant (K_D) from the plateau response at each analyte concentration. It complements SCK/MCK kinetic fitting: kinetics give k_on and k_off from the time-course, equilibrium gives K_D from the height of the plateau. The two should agree when binding is well-behaved 1:1 — disagreement is diagnostic. See BIA → Equilibrium for the underlying physics.

When to choose Equilibrium over kinetic analysis

• You only need affinity (K_D), not residence time or rate constants.
• Kinetic fitting is unstable on your data — for example mass-transport limited association, complex curve shapes, or surface decay.
• You want a robust cross-check against a kinetic K_D obtained from the same trace.

Data requirements

• A single phase with at least two ASSOCIATION injections of the same analyte at different concentrations, or two or more grouped traces with one association each. SCK-format and MCK-format data are both accepted.
• A valid concentration must be set on each association. Injections without concentrations are silently excluded; if fewer than two remain, the evaluation aborts with an error.
• For a stable 4PL fit, aim for at least six concentrations spanning two decades and bracketing the expected K_D (≥2 below, ≥2 above). Fits with fewer points succeed numerically but become unreliable.
• A BASELINE injection before each association gives the preprocessor a clean reference for offset removal. Without one the system falls back to a derived window and emits a warning.

How to run it

1. Add the measurement to a project, or open the evaluation wizard and use Add Raw Data to Project from its Data step.
2. From the project home, open Kinetics and Affinity, then chooseEquilibrium Analysis.
3. Select eligible raw data, processed data, or an existing evaluation's inputs. Configure cleanup actions in Cleanup, choose the fitting settings in Parameters, review the setup, and choose Run.
4. Results appear in the evaluation view: per-sensor 4PL fits, aggregate K_D with confidence interval, and exportable plots.

Configuration

How K_D is calculated

For each sensor, the plateau responses at each concentration are fit with the four-parameter logistic model:

y(x) = (min − max) / (1 + (x / POI)^slope) + max

• POI (point of inflection) — reported as K_D: the concentration at half-maximum response.
• min, max — bottom and top asymptotes.
• slope — Hill slope; equals 1 for a true 1:1 Langmuir isotherm.

If the concentration range does not bracket the inflection, the fitter constrains POI to the sampled range. The returned K_D is then a boundary value, not a measurement — extend the range and re-run.

Results

• K_D, Hill slope, and min/max asymptote per sensor.
• Aggregate K_D reported as the median across sensors with a 95% confidence interval.
• Dose-response plots (per-sensor and overlay), normalized binding- trace plots per ligand, per-sensor timetrace HTML plots, PDF report, and CSV exports of the fitted parameters.

Cross-check against kinetic K_D

The same trace can be evaluated both ways. Agreement within a factor of two on the same data is a strong validation of the binding model. Disagreement is informative:

Troubleshooting

• “Insufficient data for equilibrium fitting” — fewer than two ASSOCIATION injections carried valid concentrations. Check that every association has its analyte concentration set.
• “No successful equilibrium fits” — the 4PL did not converge on any sensor. Likely causes: response amplitude near noise, non-monotonic dose-response, or all concentrations on a single asymptote.
• “Baseline inferred from association” — no BASELINE injection was found before an association; the preprocessor used the early portion of the association as a substitute. Add a short baseline before each association to eliminate the warning.
• K_D looks unreasonably round (exactly the highest or lowest concentration) — the fitter has clipped the inflection to the sampled range. Extend the concentration range.

LOD (Limit of Detection) Evaluation

The LOD evaluation determines the lowest concentration of analyte that can be reliably detected above the noise level. It runs on Titration Series and Limit of Detection phase types. For the recommended experimental setup (flows, blanks, injection volumes), see Assay Setup → Limit of Detection.

From the project home, open Limit of Detection and follow the guided setup: enter the evaluation name, tags, and description, select eligible raw data, processed data, or an existing evaluation, configure the parameters below, and review the key settings before running the evaluation. Use Add Raw Data to Project to add a measurement without leaving the Data step. Auto Data Cleanup and its available actions are grouped in the main LOD Fit card; Save Intermediate Fits is in the collapsed Advanced Parameters section. Review lists cleanup actions already present in the selected data together with the automatic actions configured for the run, showing the first three names followed by an ellipsis when necessary. Select that summary to expand or collapse the complete list.

Configuration

How the LOD is calculated

With at least three blanks, the lowest detectable signal is computed as mean(blank responses) + 3.2 × SD(blank responses). One or two blanks use conservative small-sample formulas. The LOD concentration is then read off from the linear fit of the lowest-concentration points.

Results

• LOD concentration value in nM.
• Lowest detectable signal with appropriate units.
• Linear fit parameters (slope and intercept) with confidence intervals.
• Plots showing the fit and individual sensor results, generated automatically for every evaluation.
• CSV export with all calculated values and statistics.

Concentration Prediction (Standard Curve)

Estimate unknown sample concentration from a measured standard curve, tune a live 4PL, zero-baseline 3PL, or scaled through-origin cubic fit, exclude calibration observations by individual sensor or whole injection, exclude unknown injections or sensor predictions from the result tables, inspect dilution agreement and out-of-boundary findings, and save the result as a study. The Re-evaluate icon in the Evaluations table opens the full workflow with the prior evaluation preselected for both data roles and its fitting settings restored. Saving creates a new evaluation. The full workflow has its own documentation page: Concentration Prediction (Standard Curve).