Back to search

University of Waterloo

Efficient Finite-difference Methods for Sensitivity Analysis of Stiff Stochastic Discrete Models of Biochemical Systems

Abstract

dc:description.abstract

In the study of Systems Biology it is necessary to simulate cellular processes and chemical reactions that comprise biochemical systems. This is achieved through a range of mathematical modeling approaches. Standard methods use deterministic differential equations, but because many biological processes are inherently probabilistic, stochastic models must be used to capture the random fluctuations observed in these systems. The presence of noise in a system can be a significant factor in determining its behavior. The Chemical Master Equation is a valuable stochastic model of biochemical kinetics. Models based on this formalism rely on physically motivated parameters, but often these parameters are not well constrained by experiments. One important tool in the study of biochemical systems is sensitivity analysis, which aims to quantify the dependence of a system's dynamics on model parameters. Several approaches to sensitivity analysis of these models have been developed. We proposed novel methods for estimating sensitivities of discrete stochastic models of biochemical reaction systems. We used finite-difference approximations and adaptive tau-leaping strategies to estimate the sensitivities for stiff stochastic biochemical kinetics models, resulting in significant speed-up in comparison with previously published approaches for a similar accuracy. We also developed an approach for estimating sensitivity coefficients involving adaptive implicit tau-leaping strategies. We provide a comparison of these methodologies in order to identify which approach is most efficient depending of the features of the model. These results can facilitate efficient sensitivity analysis, which can serve as a foundation for the formulation, characterization, verification and reduction of models as well as further applications to identifiability analysis.

Degree

thesis:*
Grantor dc:publisher
University of Waterloo
Year dc:date.issued
2017

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Morshed, Monjur
Advisors dc:contributor.advisor
  • Ingalls, Brian
  • Ilie, Silvana

Subjects

dc:subject × 7

Rights

Language dc:language.iso
en

Identifiers

dc:identifier.*
Handle dc:identifier.uri
http://hdl.handle.net/10012/12683

Chain of custody

source
Harvested from
University of Waterloo
Base URL
dspacemainprd01.lib.uwaterloo.ca/server/oai/request
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Morshed, Monjur. Efficient Finite-difference Methods for Sensitivity Analysis of Stiff Stochastic Discrete Models of Biochemical Systems. University of Waterloo, 2017. http://hdl.handle.net/10012/12683