| Experimental and computational techniques have been developed for unraveling the structure of transcriptional regulatory networks. In Chapter 2, a whole genome, in vivo approach is presented that identifies protein-binding sites within the Escherichia coli chromosome and quantitates their occupancies. Twenty-four of the 25 sites identified in this study fell within upstream non-coding regions, with more than half falling within well-characterized cis-regulatory elements. A hierarchical clustering algorithm was used to organize the protected sites into protection clusters based on the dynamics of their occupancies across multiple growth conditions and genetic backgrounds. The membership of a protected site in a protection cluster, together with its sequence allowed us to assign a putative trans-regulatory partner to the site. In this way, our approach generates testable hypotheses for the participation of putative ORFs within well-characterized regulons. In Appendix A, we present evidence that high-density oligonucleotide arrays can be used to identify many in vivo protected sites in parallel. This approach increases both the resolution and coverage of protein binding sites and promises to be useful in studying chromatin structure in eukaryotes.;In Chapter 3, an iterative optimization-based clustering strategy was used to determine the higher order organization of the transcriptional regulatory network in yeast. Time-series of mRNA abundance, measured over two synchronized Saccharomyces cerevisiae cell cycles, were used to group 3000 genes into 30 expression clusters. The members within a cluster showed similar temporal profiles and were highly enriched for genes involved in similar functions (e.g. replication, translation or sulfur metabolism). An unbiased DNA-sequence motif search upstream of genes within a cluster identified binding sites for transcription factors with established roles in controlling many of the genes in that cluster. We also discovered many new statistically significant motifs, common to functionally homogenous members of many clusters. The identification of many expected cis-regulatory elements, together with the expression-class specificity of many novel motifs, makes this combination approach promising for the rapid elucidation of regulatory network architecture in the myriad of organisms which are now completely described at the DNA sequence level. |
