Statistics Seminar: Sue (Susan R.) Wilson -- Perspectives on Statistical Approaches to Gene Expression Data

Before undertaking any statistical data analysis, it is fundamental to understand the
nature of the measurements being made.  For gene expression microarrays, it is not
widely appreciated that the measurements are fluorescent intensities while the
biological investigations are concerned with target concentrations.  What is the
relationship between these intensities and corresponding concentrations? The first part
of the talk will explore this question, using data from spike-in experiments.  Next, to
provide insight into complex data and the resultant analyses, graphical displays can
play a useful, central role.  In particular, visualisation of gene expression profiles
can potentially contribute to the identification of co-regulated genes and gene
function.  This will be illustrated using a well-studied leukaemia data set.  Finally,
the use of gene expression data to find "signature/s" that predict disease progression
and/or outcome will be discussed using a well-known breast cancer data set, and some
novel approaches to finding stable signatures will be proposed.