Assays for recombinant proteins: a problem in non-linear calibration

Area of Science:

• Biostatistics
• Bioassay
• Protein Quantification

Background:

• Protein quantification in biological matrices (serum, plasma) often uses immunoassays or bioassays.
• Accurate analyte concentration estimation relies on non-linear calibration curves fitted to known standards.
• Existing methods face challenges in characterizing intra-assay variation and optimizing calibration.

Purpose of the Study:

• To present a general framework for calibration inference using non-linear mixed effects models.
• To address accurate characterization of intra-assay variation in bioassays.
• To evaluate the utility of empirical Bayes methods for enhancing calibration efficiency.

Main Methods:

• Development of a non-linear mixed effects model framework for calibration.
• Analysis of intra-assay variability by pooling information across multiple assay runs.
• Application of empirical Bayes methods to improve calibration curve fitting.
• Illustration using a cell-based bioassay for recombinant human relaxin.

Main Results:

• Proper characterization of intra-assay variability necessitates pooling data from multiple assay runs.
• Empirical Bayes methods demonstrate potential for significant efficiency gains in calibration.
• Practical considerations for implementing Bayesian techniques must be weighed against efficiency gains.

Conclusions:

• Non-linear mixed effects models provide a robust framework for bioassay calibration.
• Pooling assay data is crucial for accurately assessing intra-assay variability.
• Empirical Bayes methods offer a promising approach to enhance calibration efficiency in bioassays, with implementation trade-offs to consider.