Analysis of Regulatory and Interaction Networks from Clusters of Co-expressed Genes - Clustering Challenges in Biological Network - page 63

Fig. 3.5. The distribution of transcription factor connectivity from mouse vs. rat. What is evident is
the exponential distribution of transcription factor connectivity. This strongly suggests that the power
law distribution is driven by the distribution of highly selective vs. highly promiscuous transcription
factors.
overall network character does not change. Looking at the remaining connections
between the human and rat case it is apparent that even though many fewer tran-
scription factor binding sites are found at much lower frequencies, it still exhibits
a gross exponential curve.
What this suggests is that the scale free nature of transcriptional networks is
not driven primarily by the sequence of the promoter region, but rather by some
overall quality of the transcription factors. This is verified in Fig. 3.7, where a
set of randomly selected genes was subjected to the same analysis. The strong
exponential character that is present despite an essentially random sampling of
promoter regions. What is found is that many of the transcription factors are
highly promiscuous such as STAT5 and STAT6, having a consensus sequence
which is 8 bases long but only specifically recognizes 4 to 5 base pairs in the
binding sequence, which would correlate to a hit every 256 to 512 bases evaluated
respectively. Given that the promoter regions being analyzed are generally several
thousand bases, a predicted link between a gene and transcription factors such as