Abstract
Cells are, and function by, thousands of biochemical reactions and their molecules. As a whole, the system is physiologically stable, flexibly produces a large and diversified portfolio of molecules, and robustly responds to environmental and genetic perturbations. These features arise both from the kinetic and thermodynamic properties of individual molecules and from the couplings among their reactions. These couplings --- through shared intermediates, regulatory reactions, temporal stratification of rates, and many other means -- form an architecture of the system: a logical and dynamic organization of the reactions and molecules forming the cell's biochemical network. The fundamental question is how much of a network's function is due to its architecture rather than the particular properties of its components.

In this talk I summarize our work aimed at understanding the relationship between the structure and function of cellular biochemical systems. Since the scale and complexity of this problem necessitate new computational methods for biology, much of the work I will describe has been devoted to the development of these methods. To study real systems we need databases of biochemical function that are far more extensive than can be assembled by a single group. We are therefore building an electronic infrastructure for the scientific community to contribute to these databases as both depositors and reviewers, and are using this infrastructure, including a formal language of biology, to build and link the foundational databases of biochemical function. Using these preliminary data and our algorithms, we are characterizing the structure of extant biochemical architectures as a first step in understanding their integrated function. The applications of this work include the interpretation of postgenomic data obtained from microarray and proteomics experiments, identifying therapeutic strategies, predicting the metabolism of drugs and xenobiotics, and rationally designing organisms for the production of feedstock compounds for biotechnology.