Title: Statistical Analysis of Gene Expression (Microarray) Data

Ziad Taib        
Department of Biostatistics
AstraZeneca R&D Mölndal    
S-431 83, Mölndal
Sweden
e-mail Ziad.Taib@astrazeneca.com



Abstract:

Microchip arrays have become one of the most rapidly growing
techniques for monitoring gene expression at the genomic level and
thereby gaining valuable insight about various important biological
mechanisms. Examples of such mechanisms are: identifying disease
causing genes, genes involved in the regulation of some aspect of the
cell cycle, genes involved in a certain phase of the progession of a
tumour etc.

In this talk, we will discuss various statistical issues concerning
oligonucleotide microarray data. A problem that we will discuss in
some detail is that of estimating gene expression based on a proper
statistical model. More precisely, we will show how the model
introduced by (Li and Wong, 2001) can be used in its full bivariate
generality to provide a new measure of gene expression from high
density oligonucleotide arrays. We also present a second gene
expression index based on a new way of reducing the model into a
simpler univariate model. In both cases, the gene expression indices
are shown to be unbiased and to have lower variance than the
established ones.

Moreover, we present a bootstrap method aiming at providing non
parametric confidence intervals for the expression index and discuss
some additional issues concerning the statistical properties of the
different estimates.