A synopsis of some of the approaches to protein target selection is given, with an emphasis on using web resources to converge on well-ordered, readily crystallizable proteins that are maximally different from known structures. This is illustrated with the genomes of the pathogens causing tuberculosis and sleeping sickness.