The tree structure of the SCOP database proteins classification is based on four levels: families, superfamilies, folds and classes. Proteins of the same family are structurally and biologically related, with a classification guided by human expertise. Nevertheless SCOP suffers an important draw-back, which is the lack of numerical parameters characterizing each family, so as to check the possible membership to a family by using some numerical quantifications. Based on the BLOSUM Spectrum, we characterize each family by means of a cloud of points, the fingerprint of the family, that lies on a Cartesian diagram. This can be used for: 1) discriminating stronger from weaker related sequences inside the same family; 2) disclosing a fine-grained classification inside each family, based on sub-clusters sharing a very close structure; 3) classifying a novel protein on the basis of the parameters defined above.
