A method to develop images of protein families for fast comparison with any amino acid sequences is proposed.The method is based on physico-chem. properties of amino acids and on the choice of fragments of amino acid sequences of protein families that maximally distinguish the sequences from members of other families or random sequences.The efficiency of the method is demonstrated on comparison of images developed for protein families of alpha-interferons, alpha-Hbs, long neurotoxins, and DNA polymerases with the SWISS-PROT data bank.