Impact of Shorter Sequence Length on HIV-1 Drug Resistance Analysis Outcome and Clade Assignment via VirtualPhenotype

Background: VirtualPhenotype provides a quantitative assessment of anti-HIV-1 drug susceptibility from a viral genotype. The basis of this assessment is an extensive database of genotypes and corresponding drug susceptibility phenotypes obtained from HIV-1 clinical isolates. The system as developed is based on sequences encompassing the protease gene (codons 1-99) and codons 1-400 of the reverse transcriptase (RT) gene, covering all positions with known resistance-associated mutations. The purpose of the current study was to assess the impact of use of shorter sequence lengths on accuracy of phenotypic resistance analysis and on clade determination. Methods: Mutations at positions 318 and 333 of RT are known to affect susceptibility to NNRTIs and NRTIs, respectively. Full-length sequences extracted from the database were truncated to retain codons 1-99 of protease and 40 to 247 or 1 to 320 of RT, thus generating sequence lengths corresponding to those generated by other HIV-1 genotyping systems. VirtualPhenotype analyses of distinct genotypes involving positions 318 (N=98) and 333 (N=343) of RT were compared for truncated vs. full-length sequences. For assessment of clade attribution, both a new alignment tool based on ClustalW algorithm to generate alignment vs. LANL 79 HIV-1 subtype references and a new clade determination system were used. Results: Changes in susceptibility prediction (resistant to susceptible, or vice-versa) resulted when some genotypes with mutations 318 or 333 were truncated. Given the prevalence of those genotypes in the current database of clinical isolates, susceptibility prediction changed for 0.01% (efavirenz and nevirapine) to 0.19% (delavirdine) of isolates with mutation Y318F, and for 0.41% (ZDV) of isolates with mutations G333E or D. The efficacy of the enhanced clade assessment was tested by an “in silico” experiment and by testing the entire genotypic database. The effect of truncated sequence length on clade assessment was negligible (≤0.5%).