Using Radial Basis Function Networks for Parameter Estimation in Biochemical Models
Author Information
Author(s): Matsubara Yoshiya, Kikuchi Shinichi, Sugimoto Masahiro, Tomita Masaru
Primary Institution: Institute for Advanced Biosciences, Keio University
Hypothesis
Can radial basis function networks (RBFN) improve parameter estimation for stiff biochemical models compared to traditional methods?
Conclusion
The RBFN technique effectively optimized parameters for stiff biochemical models, reducing calculation time and increasing convergence rates.
Supporting Evidence
- The RBFN method reduced calculation time by more than 50%.
- The convergence rate improved from 60% to 90% with RBFN.
- Logarithmic transformation helped in simplifying the optimization process.
- ADSGA was effective in selecting additional learning data for better optimization.
Takeaway
This study shows that a special computer method can help scientists figure out important numbers for chemical reactions faster and more accurately.
Methodology
The study used radial basis function networks to estimate parameters in biochemical models, reducing the need for numerical integrations.
Potential Biases
The use of genetic algorithms may introduce sampling bias, particularly in stiff systems.
Limitations
The method may not perform as well in cases with very few time-course data sets.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website