The adenosine A3 receptor (A3R) binds adenosine and is a drug target against cancer cell proliferation. Currently, there is no experimental structure of A3R. Here, we have generated a molecular model of A3R in complex with two agonists, the nonselective 1-(6-amino-9H-purin-9-yl)-1-deoxy-N-ethyl-β-d-ribofuranuronamide (NECA) and the selective 1-deoxy-1-[6-[[(3-iodophenyl)methyl]amino]-9H-purin-9-yl]-N-methyl-β-d-ribofuranuronamide (IB-MECA). Molecular dynamics simulations of the wild-type A3R in complex with both agonists, combined with in vitro mutagenic studies revealed important residues for binding. Further, molecular mechanics-generalized Born surface area calculations were able to distinguish mutations that reduce or negate agonistic activity from those that maintained or increased the activity. Our studies reveal that selectivity of IB-MECA toward A3R requires not only direct interactions with residues within the orthosteric binding area but also with remote residues. Although V1695.30 is considered to be a selectivity filter for A3R binders, when it was mutated to glutamic acid or alanine, the activity of IB-MECA increased by making new van der Waals contacts with TM5. This result may have implications in the design of new A3R agonists.