There has been a paradigm shift in the way biology is carried out. This shift is being driven by several enabling technologies: genomics, which offers a virtually unlimited number of new gene products to study; combinatorial chemistry, which promises an almost unlimited number of compounds to act as potential agonists and antagonists; and the advent of genomewide experiments such as structural genomics, which are aimed at obtaining some functional information about as many gene products as possible. In practice, however, there are limitations to the paradigm. These limitations come in large part from the fact that the term "function" means many things, and its meaning changes depending on who is asking the question and what sorts of experiments are being employed to probe it. Genomics by itself cannot usually determine even the biochemical, much less the cellular or physiological functions of a protein. Structural biology can determine the shape of the protein but cannot reliably determine its function; the coupling between overall structure and function is a loose one. Given a structure, one cannot determine where on the surface of a protein the likely binding sites for ligands are located and what those ligands are likely to be. Genomewide experiments have many false positives and false negatives and often do not distinguish indirect effects from direct ones. The consequences of the expression of a given gene sequence can only be determined by integrating the results from many different types of experiments, and the best way to carry out this integration is not obvious.