To determine which of seven library design algorithms best introduces new protein function without destroying it altogether, seven combinatorial libraries of green fluorescent protein variants were designed and synthesized. Each was evaluated by distributions of emission intensity and color compiled from measurements made in vivo. Additional comparisons were made with a library constructed by error-prone PCR. Among the designed libraries, fluorescent function was preserved for the greatest fraction of samples in a library designed by using a structure-based computational method developed and described here. A trend was observed toward greater diversity of color in designed libraries that better preserved fluorescence. Contrary to trends observed among libraries constructed by error-prone PCR, preservation of function was observed to increase with a library's average mutation level among the four libraries designed with structure-based computational methods.GFP ͉ library design ͉ protein design ͉ protein engineering ͉ high-throughput screening P rotein sequence space is so vast that one can easily imagine the optimal sequence for a particular application will never be sampled by random mutation and recombination. Structure-based computational protein design tools seek to screen that sequence space more thoroughly than can be screened in the laboratory, but are currently based on approximate representations of candidate sequences and an incomplete understanding of the relationships between structure and function. Although many algorithms used to screen sequences in silico aim to identify a single optimal sequence (1-5), others aim instead to optimize the composition of a library of sequences (6-13). Provided that resources exist to synthesize and screen such libraries, library design algorithms compensate for the approximations built into them by increasing the number of attempts at designing the desired function. Viewed from a complementary perspective, such algorithms aim to sample sequence space more effectively than methods that randomly generate sequence diversity.Designed libraries can be synthesized for roughly the same cost as a designed sequence by recognizing the opportunities in gene synthesis for the combinatorial shuffling of sequence diversity (14-17). Although many algorithms have now been proposed to design such combinatorial libraries (7-9, 11, 12), few computationally designed libraries have been characterized experimentally (9,18,19), and, to our knowledge, there have been no controlled experiments comparing these methods with each other or with libraries of randomly generated sequence diversity. The results of such a comparison would be hard to predict, especially because none of these methods models protein function explicitly. Instead, these algorithms attempt to model protein stability as a surrogate for protein function on the assumption that libraries with a greater fraction of well folded proteins are more likely to contain variants with the desired function.Here, we evaluate seven design...