We report the use of structure-based drug design to create a selective erbB-1 (a.k.a. epidermal growth factor receptor) and erbB-2 (a.k.a. neu/her2 growth factor receptor) tyrosine kinase inhibitor. Using the X-ray crystal structure of the ternary complex of the cAMP-dependent Ser/Thr kinase together with a sequence alignment of the catalytic domains of a representative set of Ser/Thr and Tyr protein kinases, we have examined the nucleotide binding site for potential positions to attach an irreversible inhibitor. This information, combined with homology modeling of the erbB-1 and erbB-2 tyrosine kinase catalytic domains, has led to the identification of Cys797 of erbB1 and Cys805 of erbB2, which are structurally equivalent to Glu127 in the cAMP dependant Ser/Thr kinase as potential target residues. The X-ray structure of the cAMP Ser/Thr kinase shows Glu127 to be involved in a hydrogen-bonding interaction with the 2'-OH of the ribose portion of ATP. Using molecular modeling, it was predicted that the Cys side chains in erbB-1 and erbB-2 performed an analogous role, and it was postulated that the replacement of the 2'-OH of adenosine with a thiol might allow for a covalent bond to form. Since only erbB-1 and erbB-2 have a Cys at this position, the inhibitor should be selective. This model was subsequently tested experimentally by chemical synthesis of 2'-thioadenosine and assayed against the full length erbB-1 receptor and the catalytic domains of erbB-2, insulin receptor, beta-PDGF receptor, and the FGF receptor. Our results show that thioadenosine covalently inactivates erbB-1 with a second-order rate constant of k(max)/K(S) = 2000 +/- 500 M(-1) s(-1). Inactivation is fully reversed by 1 mM dithiothreitol, suggesting that inactivation involves the modification of a cysteine residue at the active site, presumably Cys797. The rate of inactivation saturates with increasing thioadenosine concentrations, suggesting that inactivation occurs through initial formation of a noncovalent complex with K(D) = 1.0 +/- 0.3 microM, followed by the slow formation of a disulfide bond with a rate constant of k(max) = (2.3 +/- 0.2) x 10(-3) s(-1). This approach may have application in the design of selective irreversible inhibitors against other members of the kinase family.