Computational design of enzymes has been a long desired capability. Large libraries of in silico proteins would be able to be screened very quickly as well as allowing the ability to design in specific interactions and residues selectively. Few examples have appeared in the literature owing to the complexity of the problem. Building on the Rosetta program which has already been shown useful for ab initio structure prediction and protein design, we have developed a general method for computational enzyme design. Starting with a description of the transition state of the reaction of interest in combination with the required catalytic residues, we first define locations in a protein scaffold, selected from the PDB, where the ligand, or transition state, can fit and make optimal geometry with the catalytic sidechains. The remainder of the active site is then designed satisfying hydrogen bonds and van der Waals packing in order to bind the substrate and to support the catalytic residues. We have made our design algorithm general in that any reaction, protein scaffold, and catalytic residues can be used. In order to demonstrate the utility of our design methodology, we selected a carbon-carbon bond cleavage reaction, the aldolase reaction, for experimentation. The aldolase mechanism we pursued includes multiple steps including a covalent intermediate. After compuatationally designing the aldolase enzymes, we synthesized, expressed, and tested the activity of the purified proteins. We have observed active enzymes using four different protein folds with rate accelerations of four orders magnitude over the original scaffold protein.