OBJECTIVES: Evaluation of the reliability of several V3-based genotypic predictors to infer viral tropism in patients infected with B and non-B strains of HIV-1. METHODS: Several genotypic tropism predictors were evaluated in plasma (RNA) samples from 198 HIV-1-infected patients, taking as gold standard the results of the phenotypic recombinant virus assay Phenoscript. In addition, for 37 B subtype HIV-1 patients the phenotypic results from plasma samples were also compared with tropism predictions based on V3 amplification from paired peripheral blood mononuclear cells (PBMCs). RESULTS: A total of 150 paired genotypic/phenotypic results were obtained from plasma specimens. Concordance values ranged from 63% to 85%, depending on the genotypic algorithm used. The best predictors in terms of sensitivity/specificity to detect X4 variants were WebPSSMX4/R5 (77%/87%), Geno2phenoFPR = 5% (80%/77%) and an algorithm combining the ‘11/25’ and ‘Net charge’ rules, termed Garridoʼs rule (80%/79%). The performance of genotypic predictors was better testing B than non-B clades. The overall sensitivity ranged from 28% to 94%, reaching 100% in subtype B antiretroviral-experienced patients using WebPSSMSI/NSI, Geno2phenoFPR ≥ 5% and Garridoʼs rule. Conversely, the sensitivity when testing non-B subtypes was poorer, ranging from 17% to 67%. Interestingly, the correlation between genotypic and phenotypic results was better when testing PBMCs than plasma using all genotypic predictors. CONCLUSIONS: Genotypic tools based on V3 sequences may provide reliable information on HIV-1 tropism when testing clade B viruses, especially in antiretroviral-experienced patients. The sensitivity to detect X4 variants using genotypic tools may improve by testing proviral DNA instead of plasma RNA.