How to Calibrate a Model

Model calibration is where a mechanistic model goes from a set of equations to a predictive tool. By fitting column and binding parameters to experimental chromatograms, the model learns how your specific molecules interact with the stationary phase.

Overview

This guide assumes you have already characterized your column — the column parameters (porosity, axial dispersion, etc) should be fixed before you start fitting binding parameters. Trying to fit everything at once is the most common source of over-parameterization.

Calibration works by comparing simulated chromatograms against experimental data and adjusting binding parameters until the simulation matches. The optimizer minimizes the difference between predicted and measured outlet concentrations across one or more experiments simultaneously.

The general workflow is:

1. Choose a binding model appropriate for your separation mode (e.g. SMA for ion exchange, HIC models for hydrophobic interaction).
2. Select 3–6 calibration experiments that cover a range of conditions — different gradient slopes, load concentrations, or salt levels.
3. Set up the optimization with reasonable parameter bounds and starting values.
4. Evaluate the fit visually and check that fitted parameters are physically realistic.
5. Validate by predicting 1–2 independent experiments that were not part of the calibration set.

A well-calibrated model should capture retention times, peak order, and approximate peak shapes across all calibration experiments. It does not need to be a perfect match — see Interpreting Results and Best Practices for guidance on what "good enough" looks like.

Choosing a Binding Model

Selecting Calibration Experiments

Setting Up the Optimization

Evaluating the Fit

Validating Against Independent Data