Stem cell differentiation study: Statistical analysis and application to microarray data

This thesis has two main aims: (1) To develop new statistical and data mining procedures for modeling stem cell genes during the differentiation process, (2) To present biological interpretations of "sternness" genes based on time-course expression patterns and genetic networks.;Given the vast information in gene expression data, it is possible to develop simple and reliable methods to investigate the differentiation of stem cells. However, very little is known about the mechanisms that characterize the genome-wide dynamic regulation of gene expression controlling stem cell differentiation. Relatively few research efforts have tackled stem cell differentiation problems using statistical procedures on time-course expression data. There is no doubt that understanding the time course of differentiation is of great significance since it should enable scientists to artificially induce stem cells to generate the appropriate intended specialized cells, such as the beating cells of the heart muscle or the insulin-producing cells of the pancreas.;In this thesis, we approach the time-course differentiation of stem cells, as measured by expression data, through some new statistical and data mining procedures. Our framework can be outlined as following: (1) Develop new detection methods for differential gene expression to identify "sternness" genes that drive and repress stem cell differentiation. (2) Develop time-course expression clustering analysis techniques and apply them to stem cells. (3) Explore new efficient analyzes for generalized linear models (GLM) that can be potentially applied in gene interaction network. Validations of computational performance are based on the experimental data available from StemBase, a well-cited public collection of DNA microarray data on stem cells designed for the purpose of facilitating stem cell research.;Keywords. stem cells, differentiation, embryonic stern cells (ESC), microarray, gene expression, clustering, gene regulatory network, time course, differential genes, Gene Ontology(GO), similarity measures, quasi-likelihood, mixture model;Stem cells have been of much interest over the past few years. Up to now, scientists have recognized many crucial features that distinguish stem cells from other types of cells: first, stem cells are capable of renewing themselves, serving as the organism's repair system. Second, they can differentiate into many types of adult cells that finally form over 200 kinds of specialized cells in the body.