Curve-Fitting Models: Linear, 4PL, 5PL, Cubic Spline

Not all assay responses are straight lines. The curve-fitting model you choose determines how accurately your calibration translates instrument readings into concentrations. Selecting an inappropriate model — applying a linear fit to an inherently sigmoidal immunoassay, for example — can introduce systematic errors that no amount of careful pipetting will correct. Understanding the assumptions and appropriate use cases for each model is therefore a core analytical competency.

Linear Regression

Linear regression (y = mx + b) is the simplest and most widely used model. It is appropriate whenever the Beer-Lambert law holds and the assay response is proportional to concentration across the working range. UV-Vis absorbance measurements of small molecules, nucleic acids (A260), and proteins (A280) typically satisfy this requirement at moderate concentrations. The K LAB Alpha and POP spectrophotometers use linear (and optionally quadratic or cubic) fitting in their Quantitation mode, achieving R² values up to 0.9999 under optimal conditions. Always confirm linearity before relying on this model — plot residuals and check that they scatter randomly around zero.

Four-Parameter Logistic (4PL)

Immunoassays — ELISAs, RIAs, and multiplex bead assays — produce sigmoidal dose-response curves that a straight line cannot describe. The four-parameter logistic model is the international standard for these applications:

y = D + (A − D) / (1 + (x / C)²)

Here A is the minimum asymptote (response at zero dose), D is the maximum asymptote, C is the inflection point (EC50), and B is the Hill slope describing curve steepness. The 4PL assumes the sigmoidal curve is symmetric around its inflection point. This is a reasonable assumption for many antibody-based assays and is the default recommendation in regulatory guidelines such as FDA Bioanalytical Method Validation guidance.

Five-Parameter Logistic (5PL)

The five-parameter logistic adds a fifth parameter (F, the asymmetry factor) that allows the curve to be asymmetric around its inflection point. This provides a better fit for assays where the upper and lower arms of the sigmoid have different shapes — common with some cytokine or hormone assays. However, 5PL fitting requires more data points to constrain all five parameters reliably and can overfit with sparse standard sets. The K LAB MRX A2000 microplate reader software supports both 4PL and 5PL, allowing the analyst to select the model that minimises residuals for their specific assay.

Cubic Spline

A cubic spline does not impose a parametric equation on the data. Instead it passes a smooth piecewise-cubic polynomial through (or near) each standard point. This makes it the most flexible option and can closely follow complex or multi-phase curves. The drawback is that spline fits do not extrapolate reliably beyond the standard range, and they can oscillate if standards are unevenly spaced. Cubic spline is best reserved for assays with dense, evenly spaced standards and well-understood response shapes.

Choosing the Right Model

• UV-Vis / colorimetric assays in the linear range — use Linear; verify with R² and residual plot.
• ELISA, immunoassay, receptor-binding assays — use 4PL as the default; switch to 5PL if residuals show systematic asymmetry.
• Non-standard or empirical curves with many standards — Cubic Spline offers maximum flexibility without imposing a mathematical form.
• Regulatory submissions — confirm the accepted model with the applicable guidance (FDA, EMA, ICH Q2); 4PL is widely accepted for ligand-binding assays.

The MRX A2000 reports goodness-of-fit metrics — R², %CV, and %RE — for every model, making it straightforward to compare fits and select the one that best describes your assay data.