A series of 48 N6-benzyladenosine 5'-uronamide derivatives has been described recently as moderately selective A3 adenosine receptor agonists of nanomolar potency (Gallo-Rodriguez, C. et al. J. Med. Chem. 1994, 37, 636). Quantitative structure activity relationships in this series, including some novel derivatives, have been investigated using a Comparative Molecular Field Analysis (CoMFA), with emphasis on the N6-substituent. The resulting three dimensional pharmacophore model defines the steric and electronic factors which modulate in vitro affinities in binding to rat brain A3 adenosine receptors. The model indicates a positive correlation of affinity with the steric characteristics of the compounds (major factor), particularly toward the 3-position of the benzyl ring of N6-benzyl NECA, and a weak correlation with the electrostatic effects of the N6-substituent. A comparison of active and inactive compounds using volume maps showed that bulk at the 3-position of the benzyl ring of the molecule is conducive to high affinity at A3 receptors, while steric bulk at other positions of the benzyl ring leads to poor binding. t-Boc-amino acid conjugates of a 3-aminobenzyl derivative were synthesized to probe the steric and hydrophobic limitations at that position. We have discovered a subregion of the N6-benzyl binding pocket occupied by a 3-(L-prolylamino) group that is sterically disallowed at A3 receptors and allowed in A1 and A2a receptors. 6-N-Phenylhydrazino and 6-O-phenylhydroxylamino derivatives, incorporating major changes in electrostatic character of the ligand proximal to the purine, were predicted by the CoMFA model to have high A3 affinity. Such analogs were synthesized and found to be well tolerated at the A3 receptor binding site.