Abstract
Objectives
To evaluate the performance of a high-throughput research assay for HIV drug resistance testing based on whole genome next-generation sequencing (NGS) that also quantifies HIV viral load.
Methods
Plasma samples (n = 145) were obtained from HIV-positive MSM (HPTN 078). Samples were analysed using clinical assays (the ViroSeq HIV-1 Genotyping System and the Abbott RealTime HIV-1 Viral Load assay) and a research assay based on whole-genome NGS (veSEQ-HIV).
Results
HIV protease and reverse transcriptase sequences (n = 142) and integrase sequences (n = 138) were obtained using ViroSeq. Sequences from all three regions were obtained for 100 (70.4%) of the 142 samples using veSEQ-HIV; results were obtained more frequently for samples with higher viral loads (93.5% for 93 samples with >5000 copies/mL; 50.0% for 26 samples with 1000–5000 copies/mL; 0% for 23 samples with <1000 copies/mL). For samples with results from both methods, drug resistance mutations (DRMs) were detected in 33 samples using ViroSeq and 42 samples using veSEQ-HIV (detection threshold: 5.0%). Overall, 146 major DRMs were detected; 107 were detected by both methods, 37 were detected by veSEQ-HIV only (frequency range: 5.0%–30.6%) and two were detected by ViroSeq only. HIV viral loads estimated by veSEQ-HIV strongly correlated with results from the Abbott RealTime Viral Load assay (R2 = 0.85; n = 142).
Conclusions
The NGS-based veSEQ-HIV method provided results for most samples with higher viral loads, was accurate for detecting major DRMs, and detected mutations at lower levels compared with a method based on population sequencing. The veSEQ-HIV method also provided HIV viral load data.
Introduction
HIV drug resistance can limit treatment options and lead to treatment failure; it can also limit the effectiveness of antiretroviral (ARV) drugs used for HIV prevention. Drug-resistant HIV can also be transmitted, limiting treatment options for drug-naive individuals.1 HIV drug resistance testing is recommended in clinical settings before initiating HIV treatment and for evaluating treatment failure.2,3 HIV drug resistance testing is also performed in research studies to inform use of ARV drugs for HIV treatment and prevention and to monitor trends in resistance in cohorts and at a population level.
HIV drug resistance testing is usually performed by sequencing defined regions of the HIV genome and identifying drug resistance mutations (DRMs). HIV genotyping is usually performed using population-sequencing methods that provide a single consensus sequence representing the viral population in a sample. These methods detect DRMs in HIV protease and reverse transcriptase (PR/RT) regions; HIV integrase is included in some assays.1Next-generation sequencing (NGS) methods provide sequence data for individual HIV viruses and are more sensitive than population-sequencing-based methods for detecting low-level DRMs.4 While minority variants with DRMs likely impact the effectiveness of ARV drugs, statistical associations with virological failure may depend on the sequencing and analysis method used.5,6
We evaluated the performance of a low-cost, high-throughput research assay that uses a virus-enrichment method to generate whole-genome HIV sequences (veSEQ-HIV); this assay also quantifies HIV viral load. This method costs approximately US$45 per sample, with a single technologist analysing approximately 360 samples per week. This method was developed to characterize HIV transmission networks but can also be used to assess HIV drug resistance and HIV viral load.7,8 The veSEQ-HIV method is described in a separate report.8 That report is focused on the performance of veSEQ-HIV for whole-genome sequencing of HIV and HIV viral load estimation in samples with non-B HIV subtypes. This report extends that study by evaluating the performance of the veSEQ-HIV assay for assessing HIV drug resistance. This report also provides information on the performance of veSEQ-HIV for analysis of samples from persons with subtype B HIV infection.
In this study, results from the veSEQ-HIV method were compared with results obtained using methods cleared by the US FDA for clinical use: the ViroSeq HIV-1 Genotyping System and the Abbott RealTime HIV Viral Load Assay.
Patients and methods
Samples used for analysis
Plasma samples were obtained from 145 HIV-positive MSM who were recruited for participation in the HIV Prevention Trials Network (HPTN) 078 study (NCT02663219). HPTN 078 was conducted in four cities in the USA (Atlanta, GA; Baltimore, MD; Birmingham, AL; and Boston, MA).9 Samples were obtained at study entry (2016–17).
Laboratory testing
HIV genotyping using the ViroSeq system was performed in a prior study at the HPTN Laboratory Center (Johns Hopkins University, Baltimore, MD, USA) using the ViroSeq HIV-1 Genotyping System, v.2.0 (Abbott Molecular, Des Plaines, IL, USA) and the ViroSeq HIV-1 Integrase Genotyping Kit, RUO (Abbott Molecular).10 This testing generated consensus sequences for HIV protease, the first 335 amino acids of HIV reverse transcriptase, and HIV integrase, and demonstrated a high frequency of DRMs in this sample set.10 HIV subtyping was also performed previously, as described.10 HIV viral load testing using the RealTime HIV-1 Viral Load Assay (Abbott Molecular) was also performed at the HPTN Laboratory Center. The limit of quantification was 40 copies/mL for this assay; some samples with limited plasma volume were tested using a validated dilution version of this assay (limit of quantification of 400 copies/mL).
Whole-genome sequencing using the veSEQ-HIV method was performed at the University of Oxford (Oxford, UK) using the Illumina MiSeq platform according to methods described previously.8 This method provides full-genome NGS sequences and quantifies HIV viral load. In short, total HIV RNA was extracted from 500 μL plasma samples; virus-specific cDNA libraries were prepared using a modified version of the SMARTer protocol (Clontech, Takara Bio, Mountain View, CA, USA), as described.8 The libraries were enriched using oligonucleotide baits and were sequenced using the Illumina MiSeq. Baits, 120 nucleotides in length and overlapping by 50 bases, were designed from a database of reference sequences from the Los Alamos National Laboratory HIV Sequences Database11 using a previously published algorithm.12 This algorithm ensures that the specificity of the baits broadly covers the known diversity of HIV-1 group M viral strains. A total of 24 PCR cycles was used in the preparation of sequencing libraries (12 cycles before bait capture and 12 cycles after bait capture). Filtered and trimmed sequences were assembled using shiver.13
A bioinformatic pipeline, drmSEQ, was devised to predict drug resistance according to the Stanford HIV Drug Resistance Database classification system based on detection of mutations in the Illumina reads generated by veSEQ-HIV. In brief, reads overlapping HIV pol genes were extracted from bam files generated by shiver, paired reads were merged, and codons were aligned using blastx with a database of 142 HIV reference sequences from the Los Alamos sequence repository.11 A multiple alignment file of the same references provided HXB2 coordinates for all mutations. Low-quality blastx alignments, alignments containing stop codons, and APOBEC (G-to-A) hypermutated sequences were removed. Samples were considered to have mutations if they were detected in at least 5% of read pairs spanning a given mutation site and a minimum of 20 read pairs in total. Mutations and combinations of mutations were scored according to the Stanford classification system (HIVdb version 8.9.1). The term ‘resistance’ used to characterize samples in this study corresponds to ‘high-level resistance’ in the Stanford system. The sequence-derived viral load was calculated from the number of de-duplicated sequencing reads obtained from each sample, using a linear regression model that was derived from samples for which independent, clinically measured viral loads were obtained. The equation was: log10(sequence-derived viral load) = 0.83 log10(number of de-duplicated reads + 1) + 1.23.8
Ethics
Written informed consent was obtained from individuals who were screened for participation in HPTN 078; this included consent for testing of stored specimens. The study was approved by institutional review boards and ethics committees at each participating institution.
Results
Samples used for analysis
Samples were obtained from HIV-positive individuals who were screened for participation in the HPTN 078 study (145 available samples; one sample per person). Testing with the ViroSeq system was successful for 142 (97.9%) of the 145 samples; 142 had results for PR/RT; 138 of those samples also had results for integrase. The median HIV viral load of the 142 samples measured with the Abbott viral load assay was 14210 copies/mL (IQR: 2507–67075 copies/mL); 138 (97.2%) of the 142 sequences were HIV subtype B; 4 samples had recombinant PR/RT sequences (subtypes B/D or B/F) with subtype B integrase sequences.
Sequences were submitted to GenBank [accession numbers: MK580177–MK580318 (HIV PR/RT) and MK580319–MK580456 (HIV integrase)].
Analysis of HIV drug resistance using veSEQ-HIV
The 142 samples described above were sequenced using veSEQ-HIV and analysed with the drmSEQ pipeline. Complete sequencing results (minimum read depth of 20) for all three regions of interest (protease, reverse transcriptase, integrase) were obtained for 100 (70.4%) of the samples using the veSEQ-HIV method. The success rate for the veSEQ-HIV method (complete sequences for all three regions) was higher for samples with higher viral loads (93.5% for the 93 samples with viral loads >5000 copies/mL; 50.0% for the 26 samples with viral loads 1000–5000 copies/mL; 0% for the 23 samples with viral loads <1000 copies/mL; 70.4% for all 142 samples). Sequencing failures with veSEQ-HIV were not biased to a particular genomic region and tended to occur in multiple regions in samples with the lowest viral loads (χ2 test of independence; P < 0.001). Testing was successful for all four samples with recombinant HIV subtypes.
HIV drug resistance results were compared for the two methods (ViroSeq and veSEQ-HIV) for the 100 samples that had complete sequencing results from both methods (Figure 1). Overall, 33.0% of the samples had resistance detected to at least one ARV drug with ViroSeq, compared with 42.0% with veSEQ-HIV (Figure 1). HIV drug resistance was detected for NNRTIs, NRTIs, PIs and integrase strand transfer inhibitors (INSTIs). Resistance to drugs in three classes was detected in a higher proportion of samples using veSEQ-HIV compared with ViroSeq (NNRTIs: 30.0% versus 23.0%; NRTIs: 19.0% versus 18.0%; INSTIs: 18.0% versus 9.0%); however, these differences were not statistically significant (P > 0.05 for all comparisons). The frequency of PI resistance was the same using both methods (4.0%). Multi-class drug resistance (resistance to two or more classes of ARV drugs) was detected in 14.0% of samples using ViroSeq and 19.0% of samples using veSEQ-HIV (Figure 1). Among the 42 samples with incomplete or no veSEQ-HIV results, 28.6% had resistance detected by the ViroSeq system.
Figure 1.
Detection of HIV DRMs using the ViroSeq system and the veSEQ-HIV method. The figure shows the percentage of individuals with resistance to at least one ARV drug (any resistance), resistance to each drug class analysed (NNRTIs, NRTIs, PIs and INSTIs), and resistance to drugs in two or more ARV drug classes (multi-class resistance). Results are shown for the ViroSeq system and the veSEQ-HIV method.
Overall, a total of 146 major DRMs were detected in the 100 samples with results from both methods; 107 were detected by both methods, and 39 were detected with one method only (Table 1). Among the 39 mutations with discordant results, 37 were detected by veSEQ-HIV only; this included 24 mutations detected at levels between 5% and 10%, 9 mutations detected at levels between 10% and 20% and 4 mutations detected at levels >20% (range: 21.2%–30.6%). In the remaining two cases, the discordant mutations were detected by ViroSeq only; in both cases, the mutations were detected at a level of <5% with veSEQ-HIV [one was detected in 8/386 reads (2.0%), and one was detected in 42/36362 reads (0.1%)].
Table 1.
Major DRMs detected
| Case | Viral load (copies/mL) | Any resistance (veSEQ/ViroSeq) | All major DRMs identified with ViroSeq and veSEQ-HIV |
Major DRMs identified with ViroSeq only | Major DRMs identified with veSEQ-HIV only | |||
|---|---|---|---|---|---|---|---|---|
| NNRTI | NRTI | PI | INSTI | |||||
| 1 | 19134 | yes/yes | K103N | M41L | T66A | INSTI: T66A (8.2%) | ||
| 2 | 163140 | yes/yes | K103S | K65R, M184V | E92Q, E138K, S147G, N155H | |||
| 3 | 254260 | yes/yes | G190A | K65R, M184V | E92Q | |||
| 4 | 174670 | yes/yes | K103N | M184V | E92Q | |||
| 5 | 11233 | yes/yes | Y188H | K65R, Y115F, M184V | T66A, E92Q, S147G | NNRTI: Y188H (30.6%), INSTI: T66A (11.8%) | ||
| 6 | 24121 | yes/yes | K103N | M184V | T66A | NRTI: M184V (8.9%), INSTI: T66A (5.6%) | ||
| 7 | 27798 | yes/yes | Y181I, G190A | Y115F, Q151M, M184V | V32I, M46L, I54L, I84V | T66A, E138K, Y143C, S147G | ||
| 8 | 197500 | yes/yes | K101E, Y181C, G190A | K65R, D67N, Q151M, K219E | V32I, M46I, M46L, I47V, I54M, V82A, I84V, L90M | Y143C | NRTI: D67N (12.9%), INSTI: Y143C (5.4%) | |
| 9 | 40530 | yes/yes | D67N, K70R, M184V, K219Q | E92Q | ||||
| 10 | 244050 | yes/yes | M184V | E92Q | ||||
| 11 | 17900 | yes/yes | M41L, M184V, T215Y | E138K, S147G, Q148R | NRTI: M41L (8.4%) | |||
| 12 | 4651 | yes/yes | M184V | G140S, Q148H | ||||
| 13 | 19002 | yes/yes | K103N | |||||
| 14 | 25650 | yes/yes | K103N | |||||
| 15 | 25830 | yes/yes | K103N | M41L | NRTI: M41L (6.4%) | |||
| 16 | 362620 | yes/yes | K103N | |||||
| 17 | 59587 | yes/yes | K103N | |||||
| 18 | 24960 | yes/yes | K103N | |||||
| 19 | 737741 | yes/yes | K103N | |||||
| 20 | 9930 | yes/yes | K103N | M41L | NRTI: M41L (14.3%) | |||
| 21 | 8780 | yes/yes | K103N | |||||
| 22 | 31027 | yes/yes | K103N | |||||
| 23 | 3231 | yes/yes | K103N | |||||
| 24 | 47740 | yes/yes | K103N | M41L | NRTI: M41L (7.8%) | |||
| 25 | 9470 | yes/yes | K103N, K103S | NNRTI: K103S (7.4%) | ||||
| 26 | 80760 | yes/yes | Y115F, M184V, M184I | NRTI: Y115F (5.1%), M184I (13.8%) | ||||
| 27 | 31050 | yes/yes | L100I | L74I, M184I | ||||
| 28 | 78498 | yes/yes | K103N | L74I, M184V, M184I | NRTI: M184I | |||
| 29 | 122990 | yes/yes | Y188L | D67N, K70R, M184V, T215F, K219E | L90M | |||
| 30 | 23546 | yes/yes | M184V | |||||
| 31 | 109399 | yes/yes | M184V | |||||
| 32 | 4571 | yes/yes | M41L, K70E, M184V, L210W, T215Y | V32I, M46I, I47V, I54L, V82T, L90M | ||||
| 33 | 14360 | yes/no | T66A | INSTI: T66A (7.6%) | ||||
| 34 | 4440 | yes/no | T66A | INSTI: T66A (5.1%) | ||||
| 35 | 7060 | yes/no | E92Q | INSTI: E92Q (6.5%) | ||||
| 36 | 1260 | yes/no | T66A | INSTI: T66A (21.2%) | ||||
| 37 | 6750 | yes/no | G190A | T66A | NNRTI: G190A (5.8%), INSTI: T66A (12.8%) | |||
| 38 | 10350 | yes/no | L74I | T66A | NRTI: L74I (10.8%), INSTI: T66A (5.9%) | |||
| 39 | 57608 | yes/no | G190A | NNRTI: G190A (6.2%) | ||||
| 40 | 37687 | yes/no | G190E | NNRTI: G190E (5.0%) | ||||
| 41 | 2890 | yes/no | G190E | NNRTI: G190E (29.3%) | ||||
| 42 | 40291 | yes/no | K101E, K103N, G190A | NNRTI: K101E (16.2%), K103N (15.9%), G190A (12.6%) | ||||
| 43 | 3758 | no/yes | M184I | I50V | NRTI: M184I | PI: I50V (6.6%) | ||
| 44 | 8200 | no/no | M41L, L210W | NRTI: L210W (24.5%) | ||||
| 45 | 2430 | no/no | D67N, K219Q | |||||
| 46 | 55100 | no/no | K70E | NRTI: K70E (5.9%) | ||||
| 47 | 11700 | no/no | I47V | PI: I47V (6.5%) | ||||
| 48 | 24450 | no/no | M41L | NRTI: M41L (5.3%) | ||||
| 49 | 634690 | no/no | M41L | NRTI: M41L (7.7%) | ||||
| 50 | 70450 | no/no | M41L | NRTI: M41L (8.7%) | ||||
| 51 | 9800 | no/no | M46I | PI: M46I (5.2%) | ||||
| 52 | 33940 | no/no | M41L | NRTI: M41L (5.6%) | ||||
Major DRMs were detected in 52 (52.0%) of 100 samples that had results from both methods (veSEQ-HIV and ViroSeq). Of those, 43 (43.0%) had high-level resistance to one or more drugs (any resistance). Overall, 107 major DRMs were detected by both methods, 37 major DRMs were detected by veSEQ-HIV only, and two major DRMs were detected by ViroSeq only (Cases 28 and 43); in both cases, the M184I mutation was detected below the 5% threshold using the veSEQ-HIV method [in 8/386 (2.0%) reads and 42/36362 (0.1%) reads]. The viral load data shown in the table were obtained at the HPTN Laboratory Center (LC) using the Abbott RealTime HIV-1 Viral Load Assay.
Analysis of HIV viral load using the veSEQ-HIV assay
Viral loads generated using the veSEQ-HIV assay were compared with viral loads obtained using the gold standard Abbott viral load assay. For the 142 samples with viral load results from both methods, the correlation of log10 viral load was 0.85 (Figure 2).
Figure 2.
Comparison of log10 viral load data obtained using a clinical viral load assay and the veSEQ-HIV method. The figure shows the comparison of log10 viral load data obtained using the Abbott viral load assay (RealTime HIV-1 Viral Load Assay, Abbott Molecular) and the log10 viral load obtained using the veSEQ-HIV method. Samples with undetectable viral load were assigned a value at the mid-point between 0 and the lower-limit of detection of the assay.
Discussion
This study evaluated the performance of the high-throughput veSEQ-HIV assay and drmSEQ analysis method for assessing drug resistance and HIV viral load. Complete sequences were obtained for ∼70% of the samples that had results from the ViroSeq system and results were obtained more frequently for samples with higher viral loads. In a previous study that evaluated >1600 sequences from Zambia, complete whole-genome sequences were obtained for 91% of the samples that had viral loads >1000 copies/mL.8
As expected, the NGS-based veSEQ-HIV method was more sensitive for detecting low-level resistant variants compared with the ViroSeq method. Overall, 42.0% of the samples were classified as resistant using veSEQ-HIV compared with 33.0% using ViroSeq. The portion of samples with major DRMs detected was substantially higher for veSEQ-HIV for NNRTIs and INSTIs. The clinical impact of low-level DRMs depends on the type of mutations detected and the ARV drug class.5,14 In this study, the low-level INSTI DRMs detected using veSEQ-HIV included mutations that conferred high-level resistance to elvitegravir [T66A (n = 9) and E92Q (n = 1)] and raltegravir [Y143C (n = 1)]. A clear association between low-level INSTI mutations and virological failure has not been established.14,15 The major DRMs detected at ≥5% with veSEQ-HIV were the same as those detected with the ViroSeq system in 107 (98.2%) of 109 cases; in the remaining two cases, the two major DRMs were detected by veSEQ-HIV at lower levels.
The veSEQ-HIV method was also used to determine HIV viral load. A correlation coefficient of R2 = 0.85 was obtained for the 142 samples tested with both methods. Of note, a correlation coefficient of 0.89 was obtained when the veSEQ-HIV method was compared with the Roche AmpliPrep TaqMan system (Roche Molecular Systems, Branchburg, NJ, USA) in a previous study that included 146 samples.8
The veSEQ method was designed and optimized for epidemiological studies so that whole-genome genealogies could be inferred with high resolution, and so that within-host diversity could be used to infer recency of infection and identify likely transmission events.8,16 The sensitivity of veSEQ-HIV for obtaining complete drug resistance data is lower than the sensitivity of most HIV genotyping assays developed for clinical use, which use targeted amplification for specific HIV regions (e.g. pol, integrase). An advantage of using whole-genome NGS in veSEQ-HIV is that the method provides quantitative information on DRMs. Further optimization of veSEQ-HIV could improve performance of the assay for samples with viral loads <5000 copies/mL and for low-volume samples that may also have limited HIV RNA even if the viral load is high. The analysis presented here demonstrates a high degree of concordance in the DRMs detected by veSEQ-HIV and a fully validated clinical HIV genotyping method. The analysis also demonstrates the expected increased sensitivity for detection of minority DRMs when viral loads are high. Additional assay validation, use of internal quantitative controls, and use of standardized reagents for assay quality assurance would be required to implement veSEQ-HIV as a clinical assay.
Consistent with a previous multicentre comparative study of bait capture sequencing,17 the quantitative relationship between read number and viral load is preserved even at low viral loads, which is evidence that contamination between high viral load and low viral load samples does not occur in sufficient quantities to distort this quantitative relationship. Despite this, very low-level sequencing artefacts cannot be excluded. A lower-bound threshold for identifying DRMs at a read depth of 20 or above is conservative since a median read depth of 20 across the whole genome corresponds to a read count of approximately 1000 reads.
In summary, veSEQ-HIV is a sensitive, high-throughput research assay for HIV genotyping and viral load quantification. This method is well suited for use in population-level surveillance and research studies evaluating interventions for HIV prevention and treatment. Other methods, or further refinements of the veSEQ-HIV method, may be needed to characterize HIV drug resistance data for samples with lower viral loads.
Acknowledgements
The authors thank the HPTN 078 study team and participants for providing the samples and data used in this study. We also thank the laboratory staff who helped with sample management and testing.
Funding
This work was supported by the HIV Prevention Trials Network (HPTN) sponsored by the National Institute of Allergy and Infectious Diseases (NIAID), National Institute on Drug Abuse (NIDA), and Office of AIDS Research, of the National Institutes of Health (NIH) [UM1-AI068613 (Eshleman); UM1-AI068617 (Donnell); and UM1-AI068619 (Cohen/El-Sadr)].
Transparency declarations
S.H.E. has collaborated on research studies with investigators from Abbott Diagnostics; Abbott Diagnostics has provided reagents for collaborative research studies. All other authors: none to declare.
Author contributions
J.M.F.: Designed the study, performed HIV genotyping, analysed data, drafted the manuscript. D.B.: Performed testing with the veSEQ-HIV system, analysed data, drafted the manuscript. V.C.: HPTN LC QAQC Representative for HPTN 078. R.B., T.G. and M. de Cesare: Assisted with development of veSEQ-HIV. E.A.W.: Statistical Research Associate for HPTN 078. T.G.: Senior Clinical Research Manager for HPTN 078. C. Del Rio, D.S.B., K.H.M. and J.E.F.: Investigator of Record for HPTN 078. J.P.H.: Statistician for HPTN 078. R.H.R.: Protocol Co-Chair for HPTN 078. C.B.: Protocol Chair for HPTN 078. C.F.: Directed testing with the veSEQ-HIV system, analysed data, drafted the manuscript. S.H.E.: Designed the study, analysed data, drafted the manuscript.
References
- 1. Clutter DS, Jordan MR, Bertagnolio S et al. HIV-1 drug resistance and resistance testing. Infect Genet Evol 2016; 46: 292–307. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Panel on Antiretroviral Guidelines for Adults and Adolescents. Guidelines for the Use of Antiretroviral Agents in Adults and Adolescents Living with HIV. Department of Health and Human Services. http://aidsinfo.nih.gov/contentfiles/lvguidelines/AdultandAdolescentGL.pdf.
- 3. Gunthard HF, Calvez V, Paredes R et al. Human immunodeficiency virus drug resistance: 2018 recommendations of the International Antiviral Society-USA Panel. Clin Infect Dis 2019; 68: 177–87. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Quinones-Mateu ME, Avila S et al. Deep sequencing: becoming a critical tool in clinical virology. J Clin Virol 2014; 61: 9–19. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Li JZ, Kuritzkes DR. Clinical implications of HIV-1 minority variants. Clin Infect Dis 2013; 56: 1667–74. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Inzaule SC, Hamers RL, Noguera-Julian M et al. Clinically relevant thresholds for ultrasensitive HIV drug resistance testing: a multi-country nested case-control study. Lancet HIV 2018; 5: e638–46. [DOI] [PubMed] [Google Scholar]
- 7. Bonsall D, Golubchik T, Kosloff B et al. HIV genotyping and phylogenetics in the HPTN 071 (PopART) study: validation of a high-throughput sequencing assay for viral load quantification, genotyping, resistance testing and high-resolution transmission networking. 22nd International AIDS Conference, Amsterdam, the Netherlands, 2018. Abstract THAA0101.
- 8. Bonsall D, Golubchik T, de Cesare M et al. A comprehensive genomics solution for HIV surveillance and clinical monitoring in low income settings. J Clin Microbiol 2020; doi:10.1128/JCM.00382-20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Remien RH, Gamble T, Farley JE et al. HPTN 078: primary results of a randomized study to engage men who have sex with men (MSM) living with HIV who are virally unsuppressed in the US. 10th IAS Conference on HIV Science, Mexico City, Mexico, 2019. Abstract MOAX0101LB.
- 10. Fogel JM, Sivay MV, Cummings V et al. HIV drug resistance in a cohort of HIV-infected MSM in the United States. AIDS 2020; 34: 91–101. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Los Alamos National Laboratory. HIV Sequence Database. http://www.hiv.lanl.gov/.
- 12. Bonsall D, Ansari MA, Ip C et al. ve-SEQ: robust, unbiased enrichment for streamlined detection and whole-genome sequencing of HCV and other highly diverse pathogens. F1000Res 2015; 13: 1062. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Wymant C, Blanquart F, Golubchik T et al. Easy and accurate reconstruction of whole HIV genomes from short-read sequence data with shiver. Virus Evol 2018; 4: vey007. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Stella-Ascariz N, Arribas JR et al. The role of HIV-1 drug-resistant minority variants in treatment failure. J Infect Dis 2017; 216 Suppl 9: S847–50. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Nguyen T, Fofana DB, Le MP et al. Prevalence and clinical impact of minority resistant variants in patients failing an integrase inhibitor-based regimen by ultra-deep sequencing. J Antimicrob Chemother 2018; 73: 2485–92. [DOI] [PubMed] [Google Scholar]
- 16. Wymant C, Hall M, Ratmann O et al. PHYLOSCANNER: inferring transmission from within- and between-host pathogen genetic diversity. Mol Biol Evol 2018; 35: 719–33. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Thomson E, Ip CL, Badhan A et al. Comparison of next-generation sequencing technologies for comprehensive assessment of full-length hepatitis C viral genomes. J Clin Microbiol 2016; 54: 2470–84. [DOI] [PMC free article] [PubMed] [Google Scholar]