1. Define the response and reported midpoint
State whether the assay measures increasing effect, remaining activity, inhibition, binding, or another endpoint. Then define whether the result is EC50, IC50, or another quantity and whether “50” means halfway between fitted asymptotes or a fixed normalized response.
This decision comes before normalization or fitting. Similar sigmoid plots can answer different biological questions, and a midpoint without its response definition is incomplete.
2. Preserve wells and apply the defined transformation
Retain raw response, plate position, sample, concentration, replicate, and control identity before deriving a normalized value. Record the blank correction or normalization formula and keep every raw component beside the result.
Decide whether fitting uses individual wells or a predefined concentration summary. A plotted mean without source wells can hide exclusions, plate position, and replicate variation.
3. Confirm that the tested range supports the midpoint
Plot every included observation against positive concentration. Confirm that the data contain informative response below and above the intended midpoint and support the plateaus required by the selected model.
A zero-dose control cannot appear on a logarithmic concentration axis. Preserve it for control assessment and handle it according to the method rather than replacing zero silently.
4. Fit the predefined model and review its evidence
Use the equation, concentration and response transformations, restrictions, weighting, bounds, and convergence rule defined by the method. Review the observed-versus-fitted plot, residuals, parameter estimates and uncertainty, bounds, warnings, and assay-specific suitability outcomes.
R² can summarize part of the fit but cannot establish supported plateaus, stable parameters, suitable residual behavior, or the biological meaning of the endpoint.
5. Report the result and its path
Report the midpoint with its response definition and concentration unit. Connect it to the source wells, transformations, model identifier and equation, parameters, exclusions and reasons, diagnostics, suitability, software version, and analysis-configuration fingerprint.
A failed or unsupported result remains failed or unsupported. Do not switch models, ranges, or exclusions after seeing the result unless the development procedure explicitly treats and retains that comparison.
Limits and model alternatives
A 4PL is a common starting point for symmetric sigmoidal behavior. A 5PL, fixed asymptote, shared parameter, or weighting scheme answers a different modeling need and requires representative evidence before routine use.
Model selection belongs in development and method approval, not in a routine attempt to improve one unattractive curve. The linked model guide addresses that decision after this source-to-result path is clear.
Frequently asked questions
Should zero concentration be included on a logarithmic x-axis?
Logarithms of zero are undefined. Preserve zero-dose controls as observations, but display or model them using an explicitly documented method rather than silently replacing zero with an arbitrary small concentration.
Is the model with the highest R² always best?
No. R² may improve when parameters are added even when the added flexibility is unstable or scientifically unhelpful. Inspect residuals, uncertainty, coverage, convergence, and repeated-run performance.