Enzymes, by far the most abundant and versatile catalysts in nature, are of enormous interest to organic chemists. High rate accelerations and exacting selectivities under mild, environmentally friendly conditions make them useful, alone or in combination with other enzymes, for the sustainable production of fine chemicals, [1][2][3][4] bioremediation, [5,6] medicine, [7,8] and diagnostic technologies. [9] Nevertheless, natural enzymes are seldom optimal for nonphysiological applications. Low stability, cosolvent sensitivity, product inhibition, and restricted substrate and reaction scope often limit their practical utility. [10] Advances in molecular biology and protein crystallography have aided efforts to finetune enzyme properties for new applications. Because it is not always possible to predict the precise outcome of protein mutation, however, evolutionary algorithms are increasingly utilized to guide enzyme optimization by a process akin to Darwinian evolution. [11][12][13] By facilitating the identification and recombination of favorable amino acid substitutions, "directed evolution" is useful for altering regio-or stereochemical outcomes, [14][15][16] enabling recognition of unnatural substrates, [17] boosting minor promiscuous activities, [18][19][20] and improving tolerance to aggressive environments. [10] Like its natural counterpart, directed evolution is comprised of three steps: creation of diversity, selection of desired phenotypes, and amplification of the selected variants. These steps can be performed iteratively, in a "steepest descent"-like optimization process, for any screenable or selectable property; no foreknowledge of structure or mechanism is required. Over the last two decades, robust methods for sequence diversification have been developed, including error-prone polymerase chain reaction (PCR) mutagenesis, [21] targeted cassette mutagenesis, [22] and DNA shuffling. [23] Coupled with powerful screening assays and selection systems, these techniques can be reliably applied to the optimization of enzymes for academic, medical, and industrial purposes. Other mutagenesis strategies, such as neutral drift [24,25] or "substrate walking", [26] have proven useful for achieving larger changes in activity or specificity. As a consequence, it is now routinely possible to adapt enzymes to a specific process of interest rather than vice versa. [2] Despite numerous examples of successful enzyme engineering, dramatic changes in function are rare and generally serendipitous. As an alternative to tweaking existing activities, de novo enzyme design is a potentially more effective strategy for producing new catalysts for completely arbitrary chemical transformations. A variety of experimental approaches have been explored toward this end. Presently, catalytic antibody technology and computational enzyme design are among the most successful (Figure 1; see Section 3.11.1). Although their impact on biocatalysis is still limited, these methods point the way for the future. This review highlights the advanta...