Statistical issues in the design and analysis of spotted microarray experiments

Michael Alan Black, Purdue University

Abstract

Spotted microarray technology is rapidly becoming a mainstay for conducting gene expression experiments. Although the ability of an array experiment to examine the expression of thousands of genes simultaneously gives a previously unheard of level of information to researchers, it also raises a plethora of statistical issues regarding both the sheer volume of data being produced, and the level of variability inherent in these relatively new array technologies. The purpose of this dissertation is to develop statistical methods for the design and analysis of spotted microarray experiments via a linear models approach. Traditional statistical concepts of experimental design, linear modeling and multiple comparisons are reviewed and considered in the context of microarray experiments. By extending and modifying techniques from these areas, a method of microarray analysis is developed that is applicable to a wide range of experimental situations. Specifically, analysis of variance methods are used in conjunction with hypothesis testing procedures to detect genes undergoing differential expression. Techniques for limiting testing errors are investigated in both the Bayesian and frequentist settings, and power calculations are proposed for determining the replication requirements of a microarray experiment. The characteristics of incomplete block designs are also examined, particularly for those experimental designs which are appropriate for microarray experiments. These analysis methods are then applied to data from Arabidopsis thaliana microarray experiments, and are shown to be successful in detecting genes exhibiting differential methylation, while providing a high level of protection against the possibility of false positive results.

Degree

Ph.D.

Advisors

Doerge, Purdue University.

Subject Area

Statistics|Molecular biology

COinS