Rapid advances in structural biology have revealed the three-dimensional structures of many biocatalysts. Molecular modeling is the tool that links these structures with experimental observations. As a qualitative tool, current modeling methods are extremely useful. They can explain, on a molecular level, unusual features of reactions. They can predict how to increase the selectivity either by substrate modification or by site-directed mutagenesis. Quantitative predictions, for example the degree of enantioselectivity, are still not reliable, however. Modeling is limited also by the availability of three-dimensional structures. Most current modeling involves hydrolases, especially proteases and lipases, but structures for other types of enzymes are starting to appear.