Engineered proteins with specificity for unnatural substrates or ligands are useful tools for studying or manipulating complex biological systems. We have engineered the prototypical tyrosine kinase v-Src to accept an unnatural ATP analogue N6-(benzyl) ATP in order to identify v-Src's direct cellular substrates. Here we have used molecular modeling to analyze the binding mode of N6-(benzyl) ATP. Based on this modeling we proposed that a new ATP analogue (N6-(2-phenethyl) ATP might be a better substrate than N6-(benzyl) ATP for the I338G mutant of v-Src. In fact the newly proposed analogue (N6-(2-phenethyl) ATP is a somewhat improved substrate for the engineered kinase (kcat = 0.6 min-1, KM = 8 microM). We also synthesized and screened three analogues of N6-(benzyl) ATP: N6-(2-methylbenzyl), ATP N6-(3-methylbenzyl), and ATP N6-(4-methylbenzyl) ATP to further probe the dimensions and shape of the introduced pocket. Results from screening newly synthesized ATP analogues agreed well with our modeling predictions. We conclude that rather than engineering a 'new' pocket by mutation of Ile 338 in v-Src to the smaller Ala or Gly residues, the I338G and I338A mutants possess a 'path' for the N6 substituent on ATP to gain access to an existing pocket in the ATP binding site. We expect to be able to extend the engineering of v-Src's ATP specificity to other kinase families based on our understanding of the binding modes of ATP analogues to engineered kinases.