The review concentrates on practical applications of computer molecular modeling in peptide drug design. The examples of the predictions (successful or not) made by computational modeling before synthesis of peptide analogs, not the explanations provided after synthesis and biological testing of peptides, are discussed. The review spans over 20 years of predictions made by computer molecular modeling for bradykinin, angiotensin, thyrotropin-releasing factor, tuftsin, substance P, CCK-related peptides, luliberin, alpha-melanotropin and opioid peptides. The described examples are discussed in terms of finding the optimal way to use computer modeling for peptide design. The step-by-step 'technology' of peptide design is outlined in detail.