Clustering Based Parameter Estimation of Thyroid Hormone Pathway.

Area of Science:

• Mathematical biology
• Endocrinology
• Computational modeling

Background:

• Thyroid hormone regulation is complex and difficult to study with limited clinical data.
• Mathematical modeling offers a powerful alternative to traditional wet lab methods for understanding thyroid function.

Purpose of the Study:

• To develop and evaluate an ordinary differential equation (ODE) model for the euthyroid thyroid system.
• To utilize genetic algorithms for subject-specific parameter estimation in thyroid function modeling.
• To enhance model accuracy by incorporating circadian variations and analyzing patient clusters.

Main Methods:

• Developed an ODE model to simulate thyroid system dynamics.
• Employed genetic algorithms for parameter estimation using available clinical data.
• Integrated a cosinor model to account for circadian rhythms in thyroid hormones.
• Applied clustering techniques to group patients with similar hormone levels for joint parameter determination.

Main Results:

• The ODE model successfully characterized thyroid system operation.
• Genetic algorithms provided subject-specific parameter estimates.
• The cosinor model significantly improved result reliability, confirmed by sensitivity analysis.
• Joint parameter determination for clustered patient groups yielded lower parameter variation compared to independent estimation.

Conclusions:

• ODE modeling combined with genetic algorithms offers a robust approach for individual thyroid function characterization.
• Incorporating circadian variations and patient clustering enhances the precision of thyroid system parameter estimation.
• This computational method provides valuable insights into thyroid physiology, especially when clinical data is scarce.