HIV-1 tropism determination is necessary prior to CCR5 antagonist use as antiretroviral therapy. Genotypic prediction of coreceptor use is a practical alternative to phenotypic tests. Cell DNA and plasma RNA-based prediction has shown discordance in many studies. We evaluate paired cell and plasma either as single or replicate V3 sequences to assess prediction comparability. The HIV-1 partial env region was sequenced and tropism was predicted using geno2pheno and position-specific scoring matrices (PSSM). Nucleotide ambiguities at V3 were quantified and genetic distance (Protdist) was determined using BioEdit. Wilcoxon signed-rank test, t tests, and Spearman correlation were performed with Prism GraphPad5.0. Results are expressed as medians, with a level of significance of p<0.05, two tailed. Single (n=28) or replicate (n=26) paired cell/plasma sequences were obtained from 54 patients. Although the clonalfalse-positive rate (FPR) value from both compartments strongly correlated (r=0.86 p<0.0001), discordance in tropism prediction was observed in both singles and replicates using geno2pheno or PSSM. Applying clonalFPR(10%) 46% (25/54) were X4 tropic, with a plasma/cell discordance of 11% in singles and 23% in replicates. Genetic distance (p<0.0001) and clonalFPR value dispersion (p=0.003) were significantly higher among replicate sequences from cells. Discordance of viral tropism prediction is not uncommon and the use of replicates does not decrease its occurrence, but improves X4 sensitivity. Sequences from provirus had greater genetic distance and dispersion of clonalFPR values. This may suggest that DNA replicate assays may better represent the diversity of HIV-1 variants, but the clinical significance of these findings needs further evaluation.