Assessing the performance of QSP models: biology as the driver for validation

Area of Science:

• Pharmacology
• Computational Biology
• Systems Biology

Background:

• Model validation is crucial for quantitative model confidence.
• Quantitative systems pharmacology (QSP) lacks standardized validation processes.
• Classical statistical validation may not fully address QSP mechanistic models.

Purpose of the Study:

• To review current QSP validation concepts.
• To contrast statistical validation aims with QSP challenges.
• To define QSP validation stages for contextual application.

Main Methods:

• Literature review of QSP validation practices.
• Comparison of statistical validation (inference, pharmacometrics, machine learning) with QSP needs.
• Analysis of published QSP models to illustrate validation levels.

Main Results:

• Current QSP validation is often piecemeal.
• Mechanistic model validation requires a tailored approach.
• Different validation levels exist, applicable based on analytical context.

Conclusions:

• A nuanced, context-specific approach to QSP model validation is necessary.
• Understanding validation stages enhances confidence in QSP model application.
• Standardized yet flexible validation frameworks are needed for QSP.