A Primer on the Visualization of Microarray Data

Introduction
DNA microarrays represent a powerful technology offering unprecedented
scope for discovery (1). However, the ability to measure, in parallel, the gene
expression patterns for thousands of genes represents both the strength and
a key weakness of microarrays. One of the central challenges of functional
genomics has been to cope with the enormity of microarray data sets, and,
indeed, the usefulness of microarrays has been limited by our ability to extract
useful information from these data. In general terms, analyzing microarray
data requires a series of numerical transformations and/or fi lters intended to
extract from the data set the subset of represented genes that may be of interest.
The resulting lists generally represent genes with large variance or periodicity
within their gene expression vectors (2); high fold inductions over a time course
(3); genes that are considered signifi cant by some statistical criterion (4); or
genes that meet some other threshold, such as exceeding a given percentile rank
in the distribution of ratios (5,6). However, examining a spreadsheet of gene
names and expression ratios often provides little insight into the interesting
trends or patterns that may exist within the data. Rather, methods have been
developed for both the classifi cation and display of these data sets. Indeed,
given the non-hypothesis-driven nature of many microarray experiments, the
ability to readily visualize trends in the data assumes paramount importance.