Related Experiment Video
Updated: May 28, 2026
07:39
Tools for the Real-Time Assessment of a Pseudomonas aeruginosa Infection Model
Published on: April 6, 2021
Hybrid modeling framework for process analytical technology: application to Bordetella pertussis cultures
M von Stosch1, R Oliveria, J Peres
1LEPAE, Departamento de Engenharia Quimica, Faculdade de Engenharia, Universidade do Porto, Rua Dr. Roberto Frias s/n, 4200-465 Porto, Portugal.
Biotechnology Progress
|November 2, 2011
Summary
A new hybrid modeling method improves data analysis in process analytical technology (PAT). This approach enhances prediction accuracy for bioprocesses like Bordetella pertussis cultivation, offering greater statistical confidence than traditional methods.
Area of Science:
- Biotechnology and Bioprocess Engineering
- Chemometrics and Data Analysis
- Process Analytical Technology (PAT)
Background:
- Process Analytical Technology (PAT) emphasizes Quality by Design, online monitoring, and closed-loop control to ensure product quality.
- Increasing use of high-throughput process analyzers generates large volumes of correlated online data.
- Traditional chemometric techniques are commonly used for PAT data analysis.
Purpose of the Study:
- To introduce and evaluate a hybrid chemometric/mathematical modeling method for PAT data analysis.
- To demonstrate the advantages of this hybrid method over standard chemometric techniques.
- To apply the methodology to analyze process data from Bordetella pertussis cultivations.
Main Methods:
- Utilized a hybrid model combining macroscopic material balance equations with nonlinear partial least squares (PLS) for reaction rate modeling.
- Applied the hybrid methodology to online sensor data (near-infrared, pH, temperature, dissolved oxygen) and off-line measurements (biomass, glutamate, lactate).
- Compared the statistical confidence and prediction errors of the hybrid model against conventional PLS models.
Main Results:
- The hybrid model demonstrated significantly higher statistical confidence compared to PLS models.
- Mean squared prediction errors were substantially reduced using the hybrid approach (e.g., lactate: 0.7190 mmol/L vs. 1.0087 mmol/L).
- Analysis of loadings and scores in the hybrid model successfully extracted relevant process features, similar to PLS.
Conclusions:
- The hybrid chemometric/mathematical modeling method offers superior performance for PAT applications.
- This approach provides more reliable process monitoring and control in biopharmaceutical manufacturing.
- The methodology enables robust extraction of process insights from complex datasets.
