Analysis of Regulatory and Interaction Networks from Clusters of Co-expressed Genes - Clustering Challenges in Biological Network - page 55

ods which combine ChiP and microarray data through the inversion of regression
techniques to estimate TFAs [10-12, 14]. Statistical, regression and decompo-
sition techniques have been proposed and successfully applied to reverse engi-
neer regulatory networks [15-22]. The main goal of reverse engineering is to
identify the activation program of transcription modules under particular condi-
tions [23] so as to hypothesize how activation/deactivation of expression can be
induced/suppressed [24]. Aside from the development of descriptive models that
correlate TFA and expression of target genes, a critical question becomes how to
identify those TFs that significantly contribute to regulation and should be mod-
ulated. Along these lines Gao et al. [12] speculated that the mRNA profile of the
target gene should be similar to the reconstructed TFA for the regulating proteins,
whereas Sun et al. [22] claimed that accurate binding information should lead to
robust TFA reconstructions.
In addition to regulatory networks, it is becoming increasingly clear that in-
teraction networks (pathway and signaling) need to be combined with gene ex-
pression data in order to establish an integrated, systems-wide, view of biological
processes and response [25-32]. Furthermore, it is becoming increasingly clear
that the structure of the interactions emanating from recorded responses captures
critical properties indicative of the state of the cellular system [33, 33-42]. The
computational analysis of regulatory, signaling and interaction pathways promises
the identification of critical targets (aka ”hubs”) with the ability to maximally im-
pact the cellular response [35]. Therefore, the systematic reconstruction of the
different types of networks (regulatory, signaling, interaction) is a critical prereq-
uisite for the deciphering the mechanisms that drive cells to carry out appropriate
functions. The existence of ”hubs”, that is the existence of a small sub-set of
highly interconnected nodes, promises profound implications due to the specu-
lated nature of these nodes. The implications of the existence of such nodes has
been speculated primarily in terms of their essential function [37-43]. It is of par-
ticular interest to analyze the robustness characteristics of this type of scale-free
network structures in terms of their robustness [44].
The purpose of this paper is to present a concurrent analysis of gene expression
data in an attempt to construct interaction networks that could be used as the
template for identifying putative points of intervention, Fig. 3.1. In order to do so,
we need to understand and characterize the structure these networks and identify
the potential implications of partial, or complete, inhibition of specific nodes in
the network.