LC Simulator: Modes Manager: “Approximation” vs. “Spline” and required run count

Description
In Modes Manager (for example, Gradient Optimization), some models use polynomial approximation (e.g., quadratic: ln k' = a + bx + cx²), while others use spline models. The “Requires … runs” panel shows the minimum number of experiments needed for peak modeling without errors, but this minimum may be lower than the number of runs needed for robust modeling.

Polynomial “Approximation” models and “Spline” models handle experimental data differently, which affects the displayed “required” number of runs.

Solution

1. When using Approximated (polynomial) models:
   • The software can lower the polynomial degree if there are not enough experimental retention times for a component.
   • The “Requires … runs” value shows the theoretical minimum number of experiments that allows the model to work without errors, not the optimal number for quality modeling.
   • The approximated model fits one function over the entire optimization range, so some predicted retention times may not perfectly match experimental values.
2. When using Spline models:
   • The software may create multiple functions over different intervals of the optimization range to reproduce experimental data very accurately.
   • Spline models assume experimental data is effectively exact and attempt to match all experimental points, which can be unrealistic if data contain noise.
3. Choosing between Approximation and Spline:
   • Use Approximation for most practical work:
     • More robust against experimental noise.
     • Better suited for routine optimization.
   • Consider Spline only when:
     • You need very high fidelity to measured retention times.
     • You understand that it may overfit experimental noise.
4. If the displayed minimum required runs seems too low:
   • Treat it as a lower bound, not a recommendation.
   • Plan additional experiments to improve model reliability, especially for critical separations.