Interactions between the hydrophobic regions of a binding site and those of a complementary ligand are often observed to provide the driving force for binding. We present a new method for the analysis of hydrophobic regions in the binding site of a protein that considers not only atom type but also the nonadditive effects arising from the shape and extent of a nonpolar region. The method has been parametrized using a purpose-built genetic algorithm to optimize its ability to identify those regions that are more likely to form a strong interaction with a nonpolar ligand group. We demonstrate the ability of this method to account for changes in the shape and extent of the exposed nonpolar surface, using both artificial and protein examples. The method is also able to rationalize differences in binding affinity for ligand-protein complexes with largely hydrophobic binding sites.