Title

Development of discrimination analysis for systems biology applications

Date of Completion

January 2006

Keywords

Engineering, Chemical|Biology, Bioinformatics

Degree

Ph.D.

Abstract

Biological systems are often modeled by a set of differential equations. For any given system there can be numerous sets proposed, each with a unique characteristic that sets it apart from other models. With more than one potential way of modeling the biological system of interest, the question will arise of which model most accurately models the system. ^ Though previously used to compare models of chemical reaction kinetics, the technique of model discrimination was introduced and applied to biological systems for the first time. Twenty-one models of HIV-1 kinetics, including one created by our group, were compared to choose the most probable model for the system. In a similar study four models of glucose-insulin relationships in type 2 diabetes were evaluated for their ability to predict glucose and insulin dynamics after ingestion of a meal. The results of both analyses proved that there is no model in the literature that has both easily measured parameters and the ability to completely capture the dynamics of the relevant system. ^ Model discrimination was applied in a novel way to identify the most probable objective function for flux balance analysis. Often times maximizing growth rate is assumed to be the appropriate objective function, but there has been little research to support this assumption. After analyzing the predictions of five metabolic objective functions, it was found that minimizing the production rate of redox potential was the most probable objective function for the genome-scale metabolism of Escherichia coli growing on succinate. ^ Finally, a genome-scale model of Bacillus anthracis was created, and based on the results of the objective function comparison for E. coli, results were generated for minimizing the production rate of redox potential for B. anthracis growing in a minimal medium. Results showed that the metabolic flux distribution agreed approximately 66% of the time, which was encouraging. However, there is still a need to improve the metabolic model of B. anthracis to more accurately predict the organism's behavior. ^