However, "directed evolution" of necessity must, to be effective in a large number of cases, involve a means of skipping between the islands of functionality in protein space:
Though mutation and screening is the essence of directed evolution, there are many factors that combine to turn the process into as much of an art as a science. These considerations include initial protein choice, mutant library construction, and method of screening.(here)Initial protein choice: starting off the process at a point that is likely to get you what you want, a key location in what you hope is an archipelago of related functionality. Mutant library construction: this is actually the method for not being bogged down by the typical Darwinian problems with searching nucleotide sequence space -- the sort Michael Behe discusses at length in The Edge of Evolution. Method of screening: In natural selection, nature is a terrible screener for the once-in-100-millenia innovation. Just because an organism has a prize-winning genetic innovation, this doesn't mean that the noise in the innovation-detection system won't overwhelm it. The survival/reproductive advantage may be washed out in the collective effect on survival, not the least of which is natural danger. Walk by a tiger, and the innovation may be lost to bad luck, and there is no shortage of this sort of bad luck in the natural world. So a protein engineer must have a better detection system, in order to ensure that he recognizes what he's looking for once it appears.