Optimal design of dynamic experiments for scalar-on-function linear models with application to a biopharmaceutical...

Area of Science:

• Statistics
• Experimental Design
• Biotechnology

Background:

• Traditional experimental designs often assume fixed variable settings.
• Functional data, where variables are functions over a domain, present unique design challenges.
• Optimizing complex biological systems like bioreactors requires advanced experimental design methods.

Purpose of the Study:

• To develop a Bayesian optimal experimental design framework for experiments with functional profile variables.
• To enable efficient identification of optimal experimental designs for scalar-on-function models.
• To provide a method for controlling the complexity of functional variables and the statistical model.

Main Methods:

• A Bayesian optimal experimental design framework was developed.
• Profile variables were represented using basis expansions within a scalar-on-function linear model.
• The approach allowed for finite-dimensional representation and optimization of functional variables.

Main Results:

• The proposed method successfully finds optimal experimental designs for functional data.
• The framework allows for effective control over the complexity of profile variables and the model.
• The approach was illustrated using dynamic feeding strategies in a bioreactor system.

Conclusions:

• The developed Bayesian framework offers a robust approach for designing experiments with functional variables.
• This method enhances the efficiency and control in optimizing complex systems.
• The application in bioreactor optimization highlights the practical utility of the proposed design strategy.