Response surface designs for experiments in bioprocessing

Area of Science:

• Statistics
• Experimental Design
• Biological Sciences

Background:

• Biological industries often use response surface methodology (RSM) for process optimization.
• Biological materials introduce significant run-to-run variation, necessitating greater experimental replication compared to other engineering fields.
• Standard RSM data analysis involves variable selection and fitting polynomial models, requiring designs that facilitate near-orthogonal parameter estimation.

Purpose of the Study:

• To introduce a novel class of three-level response surface designs tailored for biological applications.
• To develop designs that enable orthogonal estimation of most model parameters, addressing challenges posed by biological variability.
• To present subset designs that offer flexibility and improved properties for complex experimental setups.

Main Methods:

• The study introduces subset designs derived from two-level factorial designs, with non-subset factors held at their middle level.
• Fractional replicates are incorporated to expand the utility of these designs, particularly for experiments with five or more factors.
• The methodology allows for the inclusion of two- and four-level factors, as well as factors with five or more levels, through modifications to subset construction.

Main Results:

• A new class of three-level response surface designs is presented, achieving orthogonal estimation for all parameters except the quadratic terms.
• These subset designs demonstrate enhanced properties beneficial for handling biological variability and complex experimental factors.
• The designs are shown to be adaptable, allowing for the incorporation of various factor levels and the use of fractional replicates.

Conclusions:

• The proposed subset designs offer a practical and efficient approach to response surface methodology in biological industries.
• These designs effectively mitigate issues related to run-to-run variation and improve the precision of parameter estimation.
• The flexibility and adaptability of these subset designs make them valuable tools for optimizing processes involving biological materials.