Bayesian data integration for quantifying the contribution of diverse measurements to parameter estimates.

Area of Science:

• Computational biology
• Systems biology
• Biophysics

Background:

• Biological models are often underdetermined due to measurement limitations.
• Mechanistic models may have parameters unconstrained by single data types.
• Integrating diverse measurements can improve model determination.

Purpose of the Study:

• To explore the feasibility of integrating multiple measurement types for computational model determination.
• To assess if combining data enhances the accuracy of biological models.
• To investigate Bayesian statistics as a framework for integrating measurements and reducing uncertainty.

Main Methods:

• Developed an ordinary differential equation model for budding yeast cell cycle regulation.
• Integrated data from 13 studies using diverse experimental techniques.
• Utilized Bayesian inference to combine measurement types and quantify uncertainty reduction.

Main Results:

• Some model parameters were constrained by relative time course mRNA expression alone.
• Other parameters required combined measurements (e.g., relative time course and absolute transcript concentration).
• Model predictions for degradation and transcription rates matched independent experimental data; translation rate prediction revealed a model deficiency.

Conclusions:

• Integrating multiple measurement types significantly improves computational model determination.
• Combining data allows for falsification of models, as demonstrated with the translation rate parameter.
• This approach enhances the accuracy and reliability of biological models.