Case Study: Clustering Gene Expression Data

Recently developed methods for monitoring genomewide mRNA expression
changes such as oligonucleotide chips (68) and cDNA microarrays (69) are
especially powerful because they allow us to monitor quickly and inexpensively
the expression levels of a large number of genes at different time points for
different conditions, tissues, and organisms. Knowing when and under what
conditions a gene or a set of genes is expressed often provides strong clues to
their biological role and function.
Clustering algorithms are used as an essential tool to analyze these data sets
and provide valuable insight into various aspects of the genetic machinery.
There are four distinct classes of clustering problems that can be formulated
from the gene expression data sets, each addressing a different biological
problem. The fi rst problem focuses on fi nding coregulated genes by grouping
genes that have similar expression profi les. These coregulated genes can be
used to identify promoter elements by fi nding conserved areas in their upstream
regions. The second problem focuses on fi nding distinctive tissue types by
grouping tissues whose genes have similar expression profi les. These tissue
groups can then be further analyzed to identify the genes that best distinguish
the various tissues. The third clustering problem focuses on fi nding common
inducers by grouping conditions for which the expression profi les of the
genes are similar. Finding such groups of common inducers will allow us to
substitute different “trigger” mechanisms that still elicit the same response
(e.g., similar drugs, or similar herbicides or pesticides). Finally, the fourth
clustering problem focuses on fi nding organisms that exhibit similar responses
over a specifi ed set of tested conditions by grouping organisms for which the
expression profi les of their genes (in an ortholog sense) are similar. This would
allow us to identify organisms with similar responses to chosen conditions
(e.g., microbes that share a pathway).