The kinetic behaviour of human thrombin has been studied with 26 tripeptidyl-p-nitroanilide substrates protected at the N terminus and with 9 unprotected ones. By the regression analysis of experimentally determined I/&,, k,,, and /cCat/Km values the individual contribution of each side chain of the various substrates to the kinetic parameters was calculated.The contributions to the kinetic parameters of the best substrates provide information about the structure of the binding site. The interaction of subsites S1 and PI, which determines primary specificity, proved to be marginal on the basis of contribution values, though it depends upon this contact whether the substrate is hydrolyzed at all. At subsite Sz proline appeared to be favourable. Subsite S3 plays an important role in efficiency. The best parameters were obtained here with the D configurations of bulky amino acid residues. The aromatic protecting groups applied did not improve the properties of substrates. BzDPhe-Pro-Arg-Nan was predicted by calculation to be better than the protected substrates assayed. The compound was synthesized and tested. Its experimentally determined l/K,,,, 55.1 mM-', was in good agreement with 50.9 mM-' found by calculation.Several serine proteases, e.g. thrombin, trypsin, plasmin and factor Xa, hydrolyze peptide bonds at the carboxylic group of basic amino acids. Their sensitivities to substrate specificity are, however, different. Among them thrombin is the most specific which, apart from a few exceptions, splits next to arginine [1,2] [9] and growth hormones [lo]. According to Muszbek [ I l l in actin it also hydrolyzes next to lysine, whereas hydrolysis does not occur next to ornithine [12].The detailed structure of the substrate-binding sites of thrombin is not yet known. The investigations on the specificity and substrate-binding sites have so far been carried out, in addition to fibrinogen, with peptide derivatives analogous to fibrinogen sequences [13,14] [2,16] and similar hydrophobic side chains, whereas results concerning the S3 substrate-binding subsite are contradictory. Scheraga holds that S3 is rather narrow; it is conceivable, though, that more than one binding domain is operative against long peptides [12].In an attempt to study the substrate-binding sites of thrombin we applied a mathematical method [19], by the aid of which, through the regression analysis of the kinetic constants measured for various tripeptidyl-p-nitroanilide substrates, the interaction between substrate side chains and the