Standard Curves & Quantification

A standard curve fits known concentrations against measured signal, then interpolates that curve to give concentrations for unknowns. The public launch supports two workflows: Standard Curve / Sandwich ELISA and BCA / Bradford Protein.

What is supported at launch

• Standard Curve / Sandwich ELISA. Unweighted linear or quadratic calibration for analytes such as cytokines, antibodies, and antigens.
• BCA / Bradford Protein. Unweighted linear or quadratic calibration from a BSA or albumin standard, with an explicit unknown sample dilution factor when one is needed.
• Endpoint Titer ELISA. Preview. Its endpoint and dilution rules are visible in the catalog while they are validated; this is not a supported standard-curve launch workflow.

Supported calibration methods use free parameters and no weighting. 4PL, weighting, fixed or bounded parameters, and inhibitor terms are outside the supported calibration scope. Check the picker's maturity label and method description before choosing an analysis.

Prepare a table

Upload a CSV, XLSX, or XLS table with one observation per row and named columns. During setup, review the worksheet, table region, and header row selected by the inspector. If the file has multiple tables or files, review each source separately; the selected source and rows are retained with the run.

Map roles explicitly. A calibration dataset needs standard concentration and standard signal columns, plus a separate unknown signal column when unknowns are present. A sample-label column helps identify the unknowns. Keep standards and unknowns separate: an unknown label is not a numeric standard concentration. BCA requires an explicit blank role. An unknown sample dilution factor corrects the unknown concentration; it is not a serial-dose step. Enter source units and review them before starting. Confirm blank treatment and dilution explicitly. Changing a unit label does not rescale the numeric values in your file.

Plate and generic grid layouts are Preview for this launch. Export a table with one observation per row and named concentration and signal columns when possible. A plate painter selection is not automatically treated as a supported calibration input.

Choosing a model

Linear

y = mx + b is a simple, interpretable calibration for a range that is adequately described by a straight line.

Quadratic

A second-order polynomial describes a smooth curved calibration when a straight line is inadequate. Higher-order polynomials and sigmoid models are not part of the supported standard-curve contract.

Reviewing unknowns

Results keep unknown samples in source order, including duplicate labels. The table shows the source row or identifier, signal when it was recorded, the fitted concentration when estimable, and the dilution factor used for the unknown correction. The summary reports how many unknowns were estimated and how many were not estimable.

A not-estimable value is shown as such with the available reason, such as no unique root or an out-of-range signal. Missing measured signal remains “Not recorded”; it is not treated as zero. An estimated value without a source unit keeps that unit unavailable rather than gaining an inferred label. Concentrations include the recorded dilution factor, and the report does not invent confidence, LOD, LLOQ, or ULOQ values.

Intervals shown for an eligible fitted curve or parameter are separate from an unknown concentration estimate. They should not be read as a confidence interval for every unknown.

Example column layout

This synthetic example uses separate standard and unknown signal columns. Review the column roles, concentration and signal units, blank treatment, and dilution factor in setup.

Standard concentration,Standard signal,Unknown sample,Unknown signal
0,0.1,,
1,0.3,,
2,0.5,,
3,0.7,,
,,Sample A,0.4
,,Sample B,0.6

Exports and replay

• CSV, XLSX, and PDF. Numeric CSV is free; the richer workbook and PDF require Pro access. These include the saved result, source and mapping details available to the run, and unknown statuses. They do not turn an unavailable estimate into a number.
• Figure replay (Python). Replays the saved fit, observations, curves, parameters, and intervals; it does not refit the original raw data.
• Reproduce analysis (ZIP). Raw-data reproduction is a separate Pro-coming-soon capability and is offered only when the current trusted calculation and retained source inputs meet its availability checks.

For input examples and the general upload flow, see Quick start. For release maturity, use the picker labels: Supported is the launch contract; Preview requires a later validation release.