| The advent of high-throughput technologies in molecular biology has enabled a new depth of insight into systems-level regulation and control. At the forefront of these emerging technologies are gene expression microarrays. The promise of microarrays' broad utility remains, at present, held back by questions about robustness and repeatability of microarray data. In addition, there remain lingering questions about the type and scope of regulation that can be uncovered from monitoring gene expression alone. In this work, we show an integrated approach to acquisition, processing, and subsequent analysis of data. We demonstrate a family of novel information-theoretic algorithms which allow global insight into transcriptional and non-transcriptional level regulation. We show that transcriptional networks can be reconstructed on a large scale from microarray data alone, and that this can be done with a high degree of precision and sensitivity in Escherichia coli, a bacterial organism for which much curated information is available. The unsupervised nature of these algorithms enables extension of our work into poorly studied organisms, requiring no more than availability of a gene chip. In addition to transcriptional regulatory information, the inferred networks provide valuable insights into the nature of some protein complexes that leave behind a transcriptional footprint. We pursue one example of such a footprint in the search of complex, multi-level regulation. Lastly, we extend our inference of static network topology into the realm of network remodeling in order to demonstrate that strong cellular perturbations, such as drugs, leave a transcriptional trace that allows accurate reconstruction of both the perturbation target and the downstream mechanism of cellular response. Taken together, these algorithms and their applications provide a framework for analysis of whole classes of biological processes, from discovery to clinical research. |
