Parameter identifiability of power-law biochemical system models

Area of Science:

• Biotechnology
• Biochemical Engineering
• Systems Biology

Background:

• Mathematical modeling is crucial for designing and optimizing bioprocesses in biotechnology.
• Canonical models, such as power-law models in Biochemical Systems Theory, offer flexibility for nonlinear behavior simulation.
• Parameter estimation via experimental data fitting (inverse modeling) is essential but remains a bottleneck.

Purpose of the Study:

• To investigate the identifiability of parameters in power-law models using dynamic data.
• To determine if parameter values can be uniquely and accurately identified from time-series data.
• To identify the root cause of difficulties in inverse modeling of biochemical systems.

Main Methods:

• Application of existing and newly developed parameter identifiability methods.
• Analysis of two power-law models representing biochemical systems.
• Evaluation of parameter identifiability using dynamic, time-series experimental data.

Main Results:

• Lack of parametric identifiability was identified as the primary reason for challenges in inverse modeling.
• The study confirmed that parameter values cannot always be uniquely or accurately determined from dynamic data.
• Identifiability issues were demonstrated in specific power-law models of biochemical systems.

Conclusions:

• The difficulty in inverse modeling of biochemical systems stems from non-identifiable parameters.
• Parameter identifiability is a critical, often overlooked, issue in canonical modeling.
• Findings for power-law models are applicable to other canonical models in biochemical system analysis.