Accurate and reliable estimation of kinetic parameters for environmental engineering applications: A global, multi...

Area of Science:

• Environmental Engineering
• Biotechnology
• Computational Biology

Background:

• Accurate bacterial growth and metabolism predictions are vital for engineered biological systems.
• Parameter estimation in unstructured kinetic models presents significant challenges, particularly in model-data calibration.
• Existing methods struggle with multivariate, sparse, noisy data, and non-linear model structures.

Purpose of the Study:

• To present a novel global, multi-objective, and fully Bayesian optimization approach for parameter estimation.
• To overcome challenges in model-data calibration for environmental engineering applications.
• To improve the accuracy and reliability of bacterial kinetic model predictions.

Main Methods:

• Developed a sequential workflow integrating single-objective, multi-objective, and Bayesian optimization.
• Utilized global optimization to define compromise solution spaces and assess convergence.
• Employed Approximate Bayesian Computation to explore parameter and model prediction uncertainty.

Main Results:

• The proposed global optimization approach demonstrated superior parameter accuracy and precision compared to standard regression routines.
• Successfully addressed issues of premature convergence and overfitting in model calibration.
• Enabled efficient convergence and thorough exploration of parameter uncertainty.

Conclusions:

• The novel Bayesian optimization workflow provides a rigorous and effective method for estimating kinetic model parameters.
• This approach enhances the reliability of predictions for engineered biological treatment and remediation systems.
• It offers a robust solution for complex parameter estimation problems in environmental engineering.