Objective: The aim of this study was to evaluate tropism prediction performances of three algorithms [geno2pheno false-positive rate 10% (G2P10), position-specific scoring matrix (PSSM) and a combination of the 11/25 and net charge rules] and to investigate the viral and host factors potentially involved in the X4 or R5 prediction in human immunodeficiency virus-1 (HIV-1) patients. Methods: Viral tropism was determined in 179 HIV-1-infected patients eligible for CCR5 antagonist therapy. HIV-1 RNA or DNA was extracted and amplified for env gp120 sequencing. In parallel, demographic, viral, immunological and clinical determinants were analyzed. Results: According to the G2P10 algorithm, 48 patients harbored X4 or X4R5 virus. The tropism prediction was concordant for 87.7 and 88.2% of samples when comparing G2P10 with PSSM or with a combination of the 11/25 and net charge rules, respectively. X4 prediction was significantly associated with more than 35 amino acids in the V3 domain (p < 0.0001) and loss of an N-linked glycosylation site (p < 0.0001). Of the factors studied, only the nadir CD4 T-cell count was significantly associated with X4 tropism (p = 0.01). Conclusion: We determined that the X4 virus detection is closely linked to the nadir CD4 T-cell count below 100 cells/mm3 that must be taken into account when considering a CCR5 antagonist therapy switch.
HIV-1 tropism prediction by the XGboost and HMM methods