Structural and practical identifiability analysis of S-system.

Area of Science:

• Systems Biology
• Computational Biology
• Biochemical Engineering

Background:

• Biological reaction networks are commonly modeled using ordinary differential equations, with S-systems being a popular subclass.
• Current S-systems identification methods often assume structural identifiability, which may not hold true for partially measured biological systems.
• Limited data trajectories provide only a partial view of system dynamics, potentially leading to non-unique model estimations.

Purpose of the Study:

• To investigate the structural and practical identifiability of S-systems in systems biology.
• To enhance the quality of model identification for biological reaction networks with limited data.
• To explore the impact of dataset size and state space coverage on model uncertainty.

Main Methods:

• Theoretical analysis of S-system identifiability under specific assumptions.
• Application of identification techniques to the yeast fermentation pathway.
• Comparative case studies using datasets with varying state trajectory sizes.

Main Results:

• The study established conditions under which S-systems are identifiable.
• Expanding the dataset to cover a larger state space was shown to reduce model uncertainty.
• Initial concentration was identified as a factor influencing practical identifiability.

Conclusions:

• S-systems can be identifiable under defined assumptions, improving model reliability.
• Data expansion is crucial for mitigating uncertainty in S-system models derived from partial measurements.
• Understanding the role of initial conditions is key to enhancing practical identifiability in systems biology modeling.