A computer model of reverse transcriptase (RT) from human immunodeficiency virus type 1 (HIV-1) was used to design thiourea compounds that were predicted to inhibit RT. The RT model was used to approximate how changes in binding pocket shape, volume and chemical properties resulting from residue mutations would affect inhibitor binding. Our lead compound, N-[2-(2,5-dimethoxyphenylethyl)]-N'-[2-(5-bromopyridyl)]-thi ourea (HI-236) was tested against clinically observed non-nucleoside inhibitor (NNI)-resistant mutated strains of HIV. HI-236 was more potent than trovirdine, MKC-442 and zidovudine against the drug-sensitive HIV-1 strain IIIB, 50-100 times more effective than delavirdine or nevirapine and twice as effective as our recently reported lead compound N-[2-(2-fluorophenethyl)]-N'-[2-(5-bromopyridyl)]-thiourea (HI-240) against the NNI-resistant Y181C mutant HIV-1 strain A17. HI-236 was highly effective against the multidrug-resistant HIV-1 strain RT-MDR containing multiple mutations involving the RT residues 74V, 41L, 106A and 215Y. In general, thiourea compounds such as HI-236 and HI-240 showed better inhibition of drug-resistant strains of HIV-1 than thioalkylbenzyl-pyrimidine compounds such as HI-280 and HI-281. The improved activity of thioureas against RT mutants is consistent with a structural analysis of the NNI binding pocket model of RT. The activity of HI-236 against RT-MDR was superior to that of other anti-HIV agents tested, in the following order, from high to low activity; HI-236 (IC50 5 nM), HI-240 (IC50 6 nM), trovirdine (IC50 20 nM), zidovudine (IC50 150 nM), MKC-442 (IC50 300 nM), delavirdine (IC50 400 nM) and nevirapine (IC50 5 microM).