In a blind test of protein-docking algorithms, six groups used differe
nt methods to predict the structure of a protein complex. All six pred
icted structures were close enough to the experimental complex to be u
seful; nevertheless, several important details of the experimental com
plex were missed or only partially predicted.