With the increasing
number of genomes being sequenced and the availability of microarray technologies,
determining the biological function of genes is becoming a much more feasible
task. Microarrays allow the simultaneous measurement of expression
levels of thousands of genes. Since changes in protein abundance
are correlated with changes in levels of mRNA, microarrays are a useful
tool for dissecting the inner workings of a cell and determining the functions
of the various genes. A basic yet important question one can ask
in a microarray experiment is which genes change across a given set of
experimental conditions. In order to answer this question, one must
consider every gene whose expression levels have been measured. This
inevitably leads to a multiple hypothesis testing problem. Therefore,
I will develop and investigate multiple hypothesis testing procedures appropriate
for determining differential gene expression. This will include a
generalization of the Family Wise Error Rate (FWER) measure, and an extension
of the False Discovery Rate (FDR) method to dependent statistics as encountered
in microarrays. I will also develop an Empirical Bayes model for
the replicated, two-sample microarray problem, incorporating the multiple
hypothesis testing procedure results. This will provide a robust
statistical method which can be used by biologists for determining which
genes change over two conditions, yielding information about the function
of the differentially expressed genes as well as the cause of the difference
in the two samples. |
