The design of peptide drugs to treat central nervous system (CNS) diseases is hampered by our ignorance of the factors that determine whether a given peptide can cross the blood-brain-barrier (BBB). We are developing an approach to this problem that combines computer-aided ligand design, parallel synthesis of peptide libraries, and biological evaluation using in vitro BBB models. We present a genetic algorithm (GA) to search for peptides that can cross the BBB. In the design and optimization of this GA we used a genetic meta-algorithm to optimize the GA parameters. The GA is validated in silico by virtual screening of a peptide library of more than 10 1 5 molecules. We used a virtual fitness function dervied from statistical analysis of the few experimental data on peptide-BBB permeability available.