Date of Award
9-1-2011
Document Type
Thesis
Degree Name
Master of Science (MS)
Department
Applied Science
First Advisor
Nidhal Bouaynaya
Abstract
Due to the dynamic nature of biological systems, gene expression data underlying temporal processes, such as the cell development or a disease progression, can exhibit significant regime changes to facilitate regulatory functions. Deciphering these dynamical changes is crucial to understanding the biological function and developing time-varying models of cellular evolution. In this thesis, we propose to detect the different dynamical regimes of gene expression profiles using statistical sequential analysis. The proposed algorithm segments the gene time-series into distinct homogeneous regimes of expression, during which the underlying dynamics of the gene can be assumed to be stationary. Finding the switch point in gene dynamics is formulated as a hypothesis testing problem. The time-series observation (or measurement) is modeled as a stochastic process. A change in the dynamics of the time-series is modeled as a change in the parameters of the stochastic process. The use of statistical sequential analysis facilitates a unified approach for detecting both abrupt and gradual transitions. The novel dynamical change detection algorithm, presented in this thesis, uses additive and non-additive models of dynamical changes. A threshold is applied to the test statistic, which responds to changes in the time-series dynamics, in order to separate the transitions. The results are validated on synthetic data.
Recommended Citation
Pahari, Jasmine, "Segmentation of Non-Stationary Gene Expression Profiles Using Statistical Sequential Analysis" (2011). Theses and Dissertations. 309.
https://research.ualr.edu/etd/309
