How Dose-Response Curve Fitting Works
Dose-response analysis quantifies the relationship between drug concentration and biological effect. The standard approach is to fit a sigmoidal (S-shaped) model to the data using nonlinear regression.
The 4-Parameter Logistic (4PL) model is the most widely used: Y = Bottom + (Top − Bottom) / [1 + (x / IC50)^HillSlope]. The four parameters are: Top (maximum response), Bottom (minimum response), IC50 (half-maximal inhibitory concentration), and Hill slope (curve steepness).
This tool uses the Levenberg-Marquardt algorithm, the same method used by GraphPad Prism and R's drc package. It iteratively adjusts all four parameters simultaneously to minimize the sum of squared residuals between the model and your data.
The IC50 is not simply read off the graph at Y=50%. It is mathematically derived from the fitted curve parameters, which accounts for noise in individual data points and produces a more accurate estimate than visual interpolation.
Designing an IC50 Experiment
A well-designed dose-response experiment uses 8–12 concentrations spanning at least 3 log units (e.g., 0.001 µM to 100 µM), with 2–3 replicates per concentration. Half-log dilution series (3.16-fold dilutions) are standard.
Critical requirements: include concentrations that fully define both the top plateau (minimal drug effect) and bottom plateau (maximal drug effect). If the curve plateaus are missing, the fitted IC50 will be unreliable.
Normalize your data: express response as percent of untreated control (Top = 100%) or percent of maximum inhibition (Bottom = 0%). Vehicle-only controls should be included at every plate.
For screening campaigns with many compounds, a 10-point half-log dilution series from 100 µM to 0.003 µM is standard. For lead optimization, use a 12-point series centered around the expected IC50, with tighter spacing (quarter-log) near the inflection.