Analysis of Regulatory and Interaction Networks from Clusters of Co-expressed Genes - Clustering Challenges in Biological Network - page 62

factors and the incoming connectivity of the selected genes. we have focused pri-
marily upon the outgoing connectivity of the transcription factors because these
represent the underlying transcriptional network.

Fig. 3.4. Computationally constructed regulatory network (genes in green, transcription factors in

blue).

The predicted transcription factor binding sites are essentially the links within
our bi-partite network. The aggregate properties of the linkages are given in
Fig. 3.5 and Fig. 3.6. The differences between the results obtained as depicted
in Fig. 3.5 and Fig. 3.6 is the extent in which phylogenetic footprinting was used.
In Fig. 3.5, phylogenetic footprinting was done on rat and mouse, two species
which are evolutionarily very similar, while in Fig. 3.6, the phylogenetic foot-
printing was done on species which were much further evolutionarily specifically
between rat and human. This analysis was done on each individual cluster sepa-
rately to determine whether or not the process of clustering had an effect upon the
generated network.
The aggregate connectivity of the input nodes i.e. the transcription factors
shows a strong exponential decay characteristic. This is in agreement with previ-
ous evaluations upon network topology which have suggested that most biolog-
ical networks are indeed scale free. What is more surprising is that even with
the removal of a large number of connections via phylogenetic footprinting, the