Related Experiment Videos
Model discrimination using data collaboration.
Ryan Feeley1, Michael Frenklach, Matt Onsum
1Department of Mechanical Engineering, University of California, Berkeley, California 94720-1740, USA. myf@me.berkeley.edu
The Journal of Physical Chemistry. A
|May 26, 2006
Summary
This study presents a data-driven method and algorithms to compare complex kinetic reaction models. The approach effectively distinguishes between models using computable measures of data mismatch, even with many uncertain parameters.
Area of Science:
- Chemical kinetics
- Computational modeling
- Systems biology
Background:
- Kinetic reaction models are crucial for understanding complex systems.
- Discriminating between large-scale models with uncertain parameters is challenging.
- Existing methods may lack robustness for high-dimensional uncertainty.
Purpose of the Study:
- To introduce a practical, data-driven method for discriminating among large-scale kinetic reaction models.
- To develop algorithms that handle significant uncertainty in model parameters.
- To apply the method to combustion and biological signaling models.
Main Methods:
- A computable measure of model/data mismatch was developed.
- Two provably convergent algorithms were introduced to address parameter uncertainty.
- The algorithms were tested on a methane combustion model and a biological signaling network.
Main Results:
- The data-driven method successfully discriminated between kinetic models.
- Algorithms demonstrated effectiveness on a model with over 100 uncertain parameters.
- The approach was validated by distinguishing between two models of a biological signaling network.
Conclusions:
- The proposed method offers a practical approach to model discrimination in complex systems.
- The developed algorithms are robust to large ranges of parameter uncertainty.
- This work provides a valuable tool for advancing kinetic modeling and systems analysis.