Current viral vectors for gene therapy have many disadvantages: they cannot efficiently infect certain types of cells, they often stimulate an immune response, and they do not typically accommodate large genes. Nay-Wei Soong and colleagues now have developed a general strategy for creating “better” viruses (Nat. Genet. 25, 436–439, 2000). Using in vitro molecular breeding to mimic genetic variation and recombination, followed by selection, they “evolved” a virus for new tropism toward a specific target cell. The researchers combined DNA encoding the envelopes from six strains of a virus in nearly one million configurations, and tested whether the recombinant viruses could infect a cell type that none of the six strains would normally infect. By repeating the infection process five times and giving the best virus ample opportunity to outcompete the others, they managed to recover a new subtype of virus that had the ability to infect new target cells efficiently. David Curiel of the University of Alabama (Birmingham, AL) comments that although the technology might have some limitations, it has great potential in “pushing the envelope” of vector development.