Classification of protein sequences by their dipeptide composition

A simple approach to scan quickly a large protein sequence databasefor homology is described. The approach used is strictly dependenton the database organization. A database has been compiled inwhich protein sequences are grouped into families of closelyrelated proteins, each family being characterized by its averagedipeptide composition. A new entry in the database can be allocatedin a family by comparing its dipeptide composition with theaverage dipeptide composition of the families.