Gene Expression Data Analysis and Modeling

The traditional approach to research in Molecular Biology has been an inherently local one, examining and collecting data on a single gene, a single protein or a single reaction at a time. This is, of course, the classical reductionist stance: to understand the whole, one must first understand the parts. Over the years, this approach has led to remarkable achievements, allowing us to make highly accurate biochemical models of such favorites as bacteriophage Lambda. However, with the advent of the "Age of Genomics" an entirely new class of data is emerging. To date, analysis of this large scale data has consisted of little more than descriptions of how many genes were previously unknown, which genes are over-or underexpressed under certain circumstances, etc. Of course, such data is a valuable resource for researchers who are focusing on individual genes. But can we really expect to construct a detailed biochemical model of, say, an entire yeast cell with some 6000 genes (only about 1000 of which were defined before sequencing started, and about 50% of which are clearly related to other known genes), by analyzing each gene and determining all the binding and reaction constants one by one? Likewise, from the perspective of drug target identification for human disease, we cannot realistically hope to characterize all the relevant molecular interactions one-by-one as a requirement for building a predictive disease model. There is a need for methods that can handle this data in a global fashion, and that can analyze such large systems at some intermediate level, without going all the way down to the exact biochemical reactions. At the very least, such an analysis could help guide the traditional pharmacological and biochemical approaches towards those genes most worthy of attention among the thousands of newly discovered genes. Ideally, a sufficiently predictive and explanatory model at an intermediate level could obviate the need for an exact understanding of the system at the biochemical level. Large scale gene expression mapping is motivated by the premise that the information on the functional state of an organism is largely determined by the information on gene expression (based on the central dogma). This process may be conceptualized as a genetic feedback network, in which information flows from gene activity patterns through a cascade of inter-and intracellular signaling functions back to the regulation of gene expression. Gene sequence information in cis regions (regulatory inputs) and protein coding regions …