Biology Reference
In-Depth Information
localization. Thus, combinatorial regulation necessitates the assignment
of genes to several context-specific and potentially overlapping modules.
In contrast, most commonly used clustering techniques yield disjoint
partitions, assigning each gene to a single cluster.
Several examples of combinatorial regulation have been discussed in
the literature. Yuh
et al
.
22
analyzed the combinatorial logic in the control
element of a sea urchin gene. We elucidated the coregulation of the
Krebs cycle in
Saccharomyces cerevisiae
and identified two subparts of the
cycle that are autonomously coregulated under different sets of condi-
tions.
7
Several examples of condition-specific regulation in yeast and the
correlation with transcription factor binding sites were given by Gasch
and Eisen.
23
Pilpel
et al
.
24,25
pursued a systematic approach to characterize
motif combinations and their synergistic effect on expression patterns at
the genomic level. Establishing computational tools that deal with con-
text-specific coexpression is particularly relevant for higher eukaryotes,
26
since it is generally expected that the degree of combinatorial regulation
is elevated in these organisms.
1.3. Coclassification of Genes and Conditions
To take these considerations into account, expression patterns must be
analyzed with respect to specific subsets; genes and conditions should be
coclassified.
7,27-32
The resulting “transcription modules” (another com-
mon term is “biclusters”) consist of sets of coexpressed genes together
with the conditions under which this coexpression is observed. The
naive approach to evaluating the expression coherence of all possible sub-
sets of genes over all possible subsets of conditions is computationally
unfeasible, and most analysis methods for large datasets seek to limit the
search space in an appropriate way. For example, Getz
et al
.
29
introduced
a variant of biclustering based on the idea to perform standard clustering
iteratively on genes and conditions. Their coupled two-way clustering
procedure is initialized by separately clustering the genes and conditions
of the full matrix. Each combination of the resulting gene and condition
clusters defines a submatrix of the expression data. Instead of consider-
ing all possible combinations, two-way clustering is then applied to all
such submatrices in the following iteration. Other biclustering methods,
