Drug resistance testing is increasingly used to guide treatment decisions in patients infected with HIV-1. A number of rules-based algorithms have been designed to predict drug resistance profiles based on the HIV-1 genotypic data. Drug-resistance mutations in 206 viral samples from protease inhibitor (PI)-experienced subjects with HIV-1 infection were assessed, and the level of susceptibility of the samples predicted using seven unique algorithms. kappa scores were used to compare agreement of results obtained using each of the predictive algorithms with the phenotypic assay results. Good overall agreement between the different algorithms and the phenotypic results was observed. Good or excellent agreement was observed between the results obtained by the predictive algorithms and the phenotypic assay results for ritonavir, indinavir, saquinavir, and nelfinavir. For amprenavir and lopinavir, there were marked differences between the different algorithms, with poor agreement (kappa < 0.40) obtained with four of the seven algorithms for amprenavir. For lopinavir, poor agreement was obtained with three of seven algorithms using the 2.5-fold biological cut-off and four of seven with the clinical cut-off of 10. Atazanavir susceptibility was evaluated for concordance among six algorithms, with a range of 23-50% of the samples maintaining susceptibility. Although this cohort of patients included many who were highly antiretroviral experienced, predictive algorithms demonstrated good agreement with phenotype for several Pls. For those where discordance among algorithms existed, further improvement will likely occur as drug resistance pathways for the more recently approved PIs are elucidated.