The basis for the high affinity and selectivity of trimethoprim [2,4-diamino-5-(3’,4‘,5’-trimethoxybenzyl)pyrimidine, TMP] and several close structural analogues is reviewed. Methoxy group substitution on the benzyl group of 2,4-dia-minobenzylpyrimidine markedly affects both Escherichia coli dihydrofolate reductase (DHFR) K_i values and in vitro antibacterial activity. TMP is several hundred-fold more potent than the unsubstituted benzylpyrimidine, and the monomethoxy and dimethoxy analogues are of intermediate activity. However, equilibrium dissociation constants determined in the absence of cofactor (NADPH) show that the binding of these diaminobenzylpyrimidines in the enzyme-inhibitor binary complex is considerably weaker and does not vary among the compounds. Thus, the TMP binding affinity of E. coli DHFR is increased by NADPH in the ternary complex, and this increased affinity (cooperativity) varies with methoxy group substitution. In contrast, mouse DHFR has a weaker binding affinity for diaminobenzylpyrimidines, and none of the analogues show strong NADPH cooperative effects. The difference in the magnitude of NADPH/TMP cooperativity between bacterial and mammalian DHFR is an important factor in selectivity. The E. coli enzyme binds TMP more avidly in binary complex, and an additional selectivity factor of 30-fold arises from differences in cooperativity. Although the X-ray crystal structures of bacterial and vertebrate DHFR have been studied extensively, no single hypothesis convincingly explains the molecular basis of TMP selectivity. However, information on the three-dimensional structure of the enzyme has been used to rationally design novel, high-affinity inhibitors.