Biology Reference
In-Depth Information
expression and protein interaction data improves the discovery of
coherent functional groups and reduces false positives. 50,51 Similarly, the
GRAM (Genetic Regulatory Modules) algorithm by Bar-Joseph et al . 38
investigates module-regulator relationships by integrating physical reg-
ulator binding data with expression profiles. The algorithm aims to
improve the reliability of binding data by allowing lower stringency for
those interactions that are supported by coexpression; likewise, it
restricts the analysis of coexpression to genes that are targets to a com-
mon set of transcription factors.
2. Modular Algorithms
2.1. Signature Algorithm
The signature algorithm 7 (see Fig. 2 for details) was designed to test
whether a set of candidate genes exhibits a coherent expression over a
subset of the microarray data, thus already taking context-specific regu-
lation into account. These test sets are constructed by integration of
additional biological data, including functional annotations and regula-
tory sequence information.
In order to provide a more global modular picture of the transcrip-
tion program, this algorithm was later extended into an iterative scheme
(the iterative signature algorithm, or ISA; see Fig. 3) that allows for an
efficient modular decomposition of large-scale expression data (typically
tens of thousands of gene probes tested over hundreds of conditions)
also in the absence of any a priori information. 34 The ISA is one of the
state-of-the-art methods for these types of data according to various per-
formance measurements, 19,20 and has been employed in numerous bio-
logical studies (e.g. Refs. 19, 21-23). Briefly, the ISA uses a set of
expression data to identify a compendium of transcription modules, con-
sisting of coexpressed genes as well as the experimental conditions for
which this coherent expression is the most pronounced.
This algorithm has the following advantages: (1) Genes and samples
can be assigned to multiple modules (while standard clustering produces
mutually exclusive units). This is well founded from the biological point
of view, because splice variants may hybridize to the same probe and the
Search WWH ::




Custom Search