Directed evolution mimics natural selection in the laboratory. It involves: (1) random mutagenesis (error-prone PCR, DNA shuffling) to create a library of enzyme variants, (2) high-throughput screening or selection to identify improved variants, and (3) use of the best variant as a template for the next round. No prior knowledge of structure is needed. It has produced enzymes with dramatically improved thermostability, organic solvent tolerance, and altered substrate specificity.