We have developed a software system, PHIPSI,on the Connection Machine that uses a
parallel algorithm to retrieve and use information from a database of 112 known protein structures
(selected from the Brookhaven ProteinDatabank) to predict the structures of other proteins.