We have developed a method which uses a series of discriminant analyses to allocate a protein sequence of unknown function to one of 28 functional groups (toxins, immunoglobulin variable regions, cytochromes c, etc.). Allocation is based on characteristics of both global and local physical properties (hydrophobicity, charge, etc.) of the animo acide sequence, and also on the appearance in the sequence of some characteristic patterns, such as repeated consecutive appearance of certain residues, or short signature peptides. A similar method can be used to predict the cellular location of a protein segment.
