Parameter estimation of microbial models using hybrid optimization methods
Development of biological models is essential as it represents and predicts complex processes within microbial cells. These models are formed by mathematical formulations that depend heavily on a set of parameters whose accuracy is often influenced by noisy and incomplete experimental data. This stu...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2013
|
| Subjects: | |
| Online Access: | http://eprints.utm.my/id/eprint/35875/1/AfnizanfaizalAbdullahPFC2013.pdf http://eprints.utm.my/id/eprint/35875/ http://dms.library.utm.my:8080/vital/access/manager/Repository/vital:70732?site_name=Restricted Repository |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Development of biological models is essential as it represents and predicts complex processes within microbial cells. These models are formed by mathematical formulations that depend heavily on a set of parameters whose accuracy is often influenced by noisy and incomplete experimental data. This study is aimed to design and develop new optimization methods that can effectively estimate these parameters by iteratively fitting the model outputs to the experimental data. To achieve this goal, two new hybrid optimization methods based on the Firefly Algorithm (FA) method are proposed. Firstly, a method using evolutionary operations from Differential Evolution (DE) method was developed to improve the estimation accuracy of the parameters. Then, a second method using Chemical Reaction Optimization (CRO) method was proposed to surmount the convergence speed problem during parameter estimation. The effectiveness of the proposed methods was evaluated using synthetic transcriptional oscillator and extracellular protease production models. Computational experiments showed that these methods were able to estimate plausible parameters which produced model outputs that closely fitted in the experimental data. Statistical validation confirmed that these methods are competent at estimating the identifiable parameters. These findings are crucial to ensure that the estimated parameters can generate predictive and sensitive model outputs. In conclusion, this study has presented new hybrid optimization methods, capable of estimating the model parameters effectively whilst taking into account noisy and incomplete experimental data. |
|---|