Hepatitis C virus (HCV) genotype 1b displays higher genetic variability of hypervariable region 1 (HVR1) than genotype 3.

Hepatitis C virus (HCV) is characterized by high genetic variability, which is manifested both at the inter-host and intra-host levels. However, its role in the clinical course of infection is less obvious. The aim of the present study was to determine the genetic variability of HCV HVR1 (hypervariable region 1) of genotype 1b and 3 in plasma of blood donors in the early seronegative stage of infection (HCV-RNA+, anti-HCV-) and in samples from chronically infected patients using next-generation sequencing. Sequencing errors were corrected, and haplotypes inferred using the ShoRAH software. Genetic diversity parameters (intra-host number of variants, number of nucleotide substitutions and diversity per site) were assessed by DNA SP and MEGA. During the early infection, the number of variants were significantly lower in subjects infected with genotype 3 than with genotype 1b (p < 0.02). Similarly, intra-host number of variants, number of nucleotide substitutions and diversity per site were lower in genotype 3 chronic infection (p < 0.0005). In addition, early infection was characterized by significantly lower HVR1 variability values (p < 0.04) when compared to chronic infection for both genotypes. It seems that the observed differences in HVR1 variability represent an inherent property of particular viral genotypes.

Viral Metagenomics in the Clinical Realm: Lessons Learned from a Swiss-Wide Ring Trial.

Shotgun metagenomics using next generation sequencing (NGS) is a promising technique to analyze both DNA and RNA microbial material from patient samples. Mostly used in a research setting, it is now increasingly being used in the clinical realm as well, notably to support diagnosis of viral infections, thereby calling for quality control and the implementation of ring trials (RT) to benchmark pipelines and ensure comparable results. The Swiss NGS clinical virology community therefore decided to conduct a RT in 2018, in order to benchmark current metagenomic workflows used at Swiss clinical virology laboratories, and thereby contribute to the definition of common best practices. The RT consisted of two parts (increments), in order to disentangle the variability arising from the experimental compared to the bioinformatics parts of the laboratory pipeline. In addition, the RT was also designed to assess the impact of databases compared to bioinformatics algorithms on the final results, by asking participants to perform the bioinformatics analysis with a common database, in addition to using their own in-house database. Five laboratories participated in the RT (seven pipelines were tested). We observed that the algorithms had a stronger impact on the overall performance than the choice of the reference database. Our results also suggest that differences in sample preparation can lead to significant differences in the performance, and that laboratories should aim for at least 5-10 Mio reads per sample and use depth of coverage in addition to other interpretation metrics such as the percent of coverage. Performance was generally lower when increasing the number of viruses per sample. The lessons learned from this pilot study will be useful for the development of larger-scale RTs to serve as regular quality control tests for laboratories performing NGS analyses of viruses in a clinical setting.

Profiling host ANP32A splicing landscapes to predict influenza A virus polymerase adaptation.

Species' differences in cellular factors limit avian influenza A virus (IAV) zoonoses and human pandemics. The IAV polymerase, vPol, harbors evolutionary sites to overcome restriction and determines virulence. Here, we establish host ANP32A as a critical driver of selection, and identify host-specific ANP32A splicing landscapes that predict viral evolution. We find that avian species differentially express three ANP32A isoforms diverging in a vPol-promoting insert. ANP32As with shorter inserts interact poorly with vPol, are compromised in supporting avian-like IAV replication, and drive selection of mammalian-adaptive vPol sequences with distinct kinetics. By integrating selection data with multi-species ANP32A splice variant profiling, we develop a mathematical model to predict avian species potentially driving (swallow, magpie) or maintaining (goose, swan) mammalian-adaptive vPol signatures. Supporting these predictions, surveillance data confirm enrichment of several mammalian-adaptive vPol substitutions in magpie IAVs. Profiling host ANP32A splicing could enhance surveillance and eradication efforts against IAVs with pandemic potential.

Metagenomic Virome Sequencing in Living Donor and Recipient Kidney Transplant Pairs Revealed JC Polyomavirus Transmission.

BACKGROUND: Before kidney transplantation, donors and recipients are routinely screened for viral pathogens using specific tests. Little is known about unrecognized viruses of the urinary tract that potentially result in transmission. Using an open metagenomic approach, we aimed to comprehensively assess virus transmission in living-donor kidney transplantation. METHODS: Living kidney donors and their corresponding recipients were enrolled at the time of transplantation. Follow-up study visits for recipients were scheduled 4-6 weeks and 1 year thereafter. At each visit, plasma and urine samples were collected and transplant recipients were evaluated for signs of infection or other transplant-related complications. For metagenomic analysis, samples were enriched for viruses, amplified by anchored random polymerase chain reaction (PCR), and sequenced using high-throughput metagenomic sequencing. Viruses detected by sequencing were confirmed using real-time PCR. RESULTS: We analyzed a total of 30 living kidney donor and recipient pairs, with a follow-up of at least 1 year. In addition to viruses commonly detected during routine post-transplant virus monitoring, metagenomic sequencing detected JC polyomavirus (JCPyV) in the urine of 7 donors and their corresponding recipients. Phylogenetic analysis confirmed infection with the donor strain in 6 cases, suggesting transmission from the transplant donor to the recipient, despite recipient seropositivity for JCPyV at the time of transplantation. CONCLUSIONS: Metagenomic sequencing identified frequent transmission of JCPyV from kidney transplant donors to recipients. Considering the high incidence rate, future studies within larger cohorts are needed to define the relevance of JCPyV infection and the donor's virome for transplant outcomes.

Low prevalence of transmitted HIV-1 drug resistance detected by a dried blood spot (DBS)-based next-generation sequencing (NGS) method in newly diagnosed individuals in Cameroon in the years 2015-16.

Objectives: To determine the most recent prevalence, transmission patterns and risk factors of transmitted drug-resistance mutations (TDRMs) in Cameroon, we initiated a multicentre study monitoring HIV-1 drug resistance in newly HIV-1-diagnosed individuals using a novel next-generation sequencing (NGS) assay applicable to fingerprick dried blood spot (DBS) samples. Methods: Fingerprick DBS samples and questionnaires were collected from 360 newly HIV-1-diagnosed individuals in four hospitals in urban areas in Cameroon in the years 2015-16. We developed an HIV-1 protease and reverse transcriptase drug resistance genotyping assay applicable to DBS samples and HIV-1 genomes of groups M, N and O. The WHO 2009 list of mutations for surveillance of transmitted drug-resistant HIV strains was used to analyse TDRMs. Results: Applying our 'DBS-NGS-genotypic resistance test', baseline HIV-1 drug resistance data were successfully obtained from 82.8% (298/360) of newly diagnosed individuals. At nucleotide frequencies >15%, TDRMs to NRTIs were observed in 3.0% (9/298), to NNRTIs in 4.0% (12/298) and to PIs in 1.3% (3/240). The NNRTI mutation K103N was most commonly detected (2.7%). Expanding the analysis to low-abundance TDRMs, i.e. 3%-15%, 12 additional individuals (4.0%) harbouring TDRMs were identified. Having unprotected sex with a known HIV-1-positive person was significantly associated with the transmission of DRMs (adjusted OR 9.6; 95% CI 1.79-51.3). Conclusions: The prevalence of transmitted HIV-1 drug resistance is currently low in the study sites in Cameroon. Evidence of some risky sexual behaviours depicts a public health problem with possible implications for the prevention of new HIV-1 infections.

Next-generation sequencing analysis of a cluster of hepatitis C virus infections in a haematology and oncology center.

Molecular characterization of early hepatitis C virus (HCV) infection remains rare. Ten out of 78 patients of a hematology/oncology center were found to be HCV RNA positive two to four months after hospitalization. Only two of the ten patients were anti-HCV positive. HCV hypervariable region 1 (HVR1) was amplified in seven patients (including one anti-HCV positive) and analyzed by next generation sequencing (NGS). Genetic variants were reconstructed by Shorah and an empirically established 0.5% variant frequency cut-off was implemented. These sequences were compared by phylogenetic and diversity analyses. Ten unrelated blood donors with newly acquired HCV infection detected at the time of donation (HCV RNA positive and anti-HCV negative) served as controls. One to seven HVR1 variants were found in each patient. Sequences intermixed phylogenetically with no evidence of clustering in individual patients. These sequences were more similar to each other (similarity 95.4% to 100.0%) than to those of controls (similarity 64.8% to 82.6%). An identical predominant variant was present in four patients, whereas other closely related variants dominated in the remaining three patients. In five patients the HCV population was limited to a single variant or one predominant variant and minor variants of less than 10% frequency. In conclusion, NGS analysis of a cluster of HCV infections acquired in the hospital setting revealed the presence of low diversity, very closely related variants in all patients, suggesting an early-stage infection with the same virus. NGS combined with phylogenetic analysis and classical epidemiological analysis could help in tracking of HCV outbreaks.

Persistent mammalian orthoreovirus, coxsackievirus and adenovirus co-infection in a child with a primary immunodeficiency detected by metagenomic sequencing: a case report.

BACKGROUND: We report a rare case of Mammalian orthoreovirus (MRV) infection in a child with a primary immunodeficiency (PID). Infections with Mammalian orthoreovirus are very rare and probably of zoonotic origin. Only a few cases have been described so far, including one with similar pathogenesis as in our case. CASE PRESENTATION: The patient, age 11, presented with flu-like symptoms and persistent severe diarrhea. Enterovirus has been detected over several months, however, exact typing of a positive cell culture remained inconclusive. Unbiased metagenomic sequencing then detected MRV in stool samples from several time points. The sequencing approach further revealed co-infection with a recombinant Coxsackievirus and Adenovirus. MRV-specific antibodies detected by immunofluorescence proved that the patient seroconverted. CONCLUSION: This case highlights the potential of unbiased metagenomic sequencing in supplementing routine diagnostic methods, especially in situations of chronic infection with multiple viruses as seen here in an immunocompromised host. The origin, transmission routes and implications of MRV infection in humans merit further investigation.

Optimization and validation of sample preparation for metagenomic sequencing of viruses in clinical samples.

BACKGROUND: Sequence-specific PCR is the most common approach for virus identification in diagnostic laboratories. However, as specific PCR only detects pre-defined targets, novel virus strains or viruses not included in routine test panels will be missed. Recently, advances in high-throughput sequencing allow for virus-sequence-independent identification of entire virus populations in clinical samples, yet standardized protocols are needed to allow broad application in clinical diagnostics. Here, we describe a comprehensive sample preparation protocol for high-throughput metagenomic virus sequencing using random amplification of total nucleic acids from clinical samples. RESULTS: In order to optimize metagenomic sequencing for application in virus diagnostics, we tested different enrichment and amplification procedures on plasma samples spiked with RNA and DNA viruses. A protocol including filtration, nuclease digestion, and random amplification of RNA and DNA in separate reactions provided the best results, allowing reliable recovery of viral genomes and a good correlation of the relative number of sequencing reads with the virus input. We further validated our method by sequencing a multiplexed viral pathogen reagent containing a range of human viruses from different virus families. Our method proved successful in detecting the majority of the included viruses with high read numbers and compared well to other protocols in the field validated against the same reference reagent. Our sequencing protocol does work not only with plasma but also with other clinical samples such as urine and throat swabs. CONCLUSIONS: The workflow for virus metagenomic sequencing that we established proved successful in detecting a variety of viruses in different clinical samples. Our protocol supplements existing virus-specific detection strategies providing opportunities to identify atypical and novel viruses commonly not accounted for in routine diagnostic panels.

Metagenomic sequencing complements routine diagnostics in identifying viral pathogens in lung transplant recipients with unknown etiology of respiratory infection.

BACKGROUND: Lung transplant patients are a vulnerable group of immunosuppressed patients that are prone to frequent respiratory infections. We studied 60 episodes of respiratory symptoms in 71 lung transplant patients. Almost half of these episodes were of unknown infectious etiology despite extensive routine diagnostic testing. METHODS: We re-analyzed respiratory samples of all episodes with undetermined etiology in order to detect potential viral pathogens missed/not accounted for in routine diagnostics. Respiratory samples were enriched for viruses by filtration and nuclease digestion, whole nucleic acids extracted and randomly amplified before high throughput metagenomic virus sequencing. Viruses were identified by a bioinformatic pipeline and confirmed and quantified using specific real-time PCR. RESULTS: In completion of routine diagnostics, we identified and confirmed a viral etiology of infection by our metagenomic approach in four patients (three Rhinovirus A, one Rhinovirus B infection) despite initial negative results in specific multiplex PCR. Notably, the majority of samples were also positive for Torque teno virus (TTV) and Human Herpesvirus 7 (HHV-7). While TTV viral loads increased with immunosuppression in both throat swabs and blood samples, HHV-7 remained at low levels throughout the observation period and was restricted to the respiratory tract. CONCLUSION: This study highlights the potential of metagenomic sequencing for virus diagnostics in cases with previously unknown etiology of infection and in complex diagnostic situations such as in immunocompromised hosts.

Delineating CD4 dependency of HIV-1: Adaptation to infect low level CD4 expressing target cells widens cellular tropism but severely impacts on envelope functionality.

A hallmark of HIV-1 infection is the continuously declining number of the virus' predominant target cells, activated CD4+ T cells. With diminishing CD4+ T cell levels, the capacity to utilize alternate cell types and receptors, including cells that express low CD4 receptor levels such as macrophages, thus becomes crucial. To explore evolutionary paths that allow HIV-1 to acquire a wider host cell range by infecting cells with lower CD4 levels, we dissected the evolution of the envelope-CD4 interaction under in vitro culture conditions that mimicked the decline of CD4high target cells, using a prototypic subtype B, R5-tropic strain. Adaptation to CD4low targets proved to severely alter envelope functions including trimer opening as indicated by a higher affinity to CD4 and loss in shielding against neutralizing antibodies. We observed a strikingly decreased infectivity on CD4high target cells, but sustained infectivity on CD4low targets, including macrophages. Intriguingly, the adaptation to CD4low targets altered the kinetic of the entry process, leading to rapid CD4 engagement and an extended transition time between CD4 and CCR5 binding during entry. This phenotype was also observed for certain central nervous system (CNS) derived macrophage-tropic viruses, highlighting that the functional perturbation we defined upon in vitro adaptation to CD4low targets occurs in vivo. Collectively, our findings suggest that CD4low adapted envelopes may exhibit severe deficiencies in entry fitness and shielding early in their evolution. Considering this, adaptation to CD4low targets may preferentially occur in a sheltered and immune-privileged environment such as the CNS to allow fitness restoring compensatory mutations to occur.

MinVar: A rapid and versatile tool for HIV-1 drug resistance genotyping by deep sequencing.

Genotypic monitoring of drug-resistance mutations (DRMs) in HIV-1 infected individuals is strongly recommended to guide selection of the initial antiretroviral therapy (ART) and changes of drug regimens. Traditionally, mutations conferring drug resistance are detected by population sequencing of the reverse transcribed viral RNA encoding the HIV-1 enzymes target by ART, followed by manual analysis and interpretation of Sanger sequencing traces. This process is labor intensive, relies on subjective interpretation from the operator, and offers limited sensitivity as only mutations above 20% frequency can be reliably detected. Here we present MinVar, a pipeline for the analysis of deep sequencing data, which allows reliable and automated detection of DRMs down to 5%. We evaluated MinVar with data from amplicon sequencing of defined mixtures of molecular virus clones with known DRM and plasma samples of viremic HIV-1 infected individuals and we compared it to VirVarSeq, another virus variant detection tool exclusively working on Illumina deep sequencing data. MinVar was designed to be compatible with a diverse range of sequencing platforms and allows the detection of DRMs and insertions/deletions from deep sequencing data without the need to perform additional bioinformatics analysis, a prerequisite to a widespread implementation of HIV-1 genotyping using deep sequencing in routine diagnostic settings.

Spouse-to-Spouse Transmission and Evolution of Hypervariable Region 1 and 5' Untranslated Region of Hepatitis C Virus Analyzed by Next-Generation Sequencing.

Hepatitis C virus (HCV) transmission between spouses remains poorly characterized, largely due to the limited availability of samples from the early stage of infection, as well as methodological constraints. A fifty-eight year-old male developed acute hepatitis C infection and his 53-year old spouse has been HCV-positive for over 10 years. Serum samples were collected from both at the time of acute hepatitis C diagnosis in male (baseline) and then at 9 and 13 months. Hypervariable region 1 (HVR1) and 5' untranslated region (5'UTR) sequences were amplified and subjected to next generation sequencing (NGS) using a pyrosequencing platform. Genetic variants were inferred by Shorah reconstruction method and compared by phylogenetic and sequence diversity analysis. As the sequencing error of the procedure was previously determined to be ≤ 1.5%, the analysis was conducted with and without the 1.5% cut-off with regard to the frequency of variants. No identical HVR1 variants were identified in spouses at baseline and follow-up samples regardless whether the cut-off was applied or not. However, there was high similarity (98.3%) between a minor baseline donor variant (1.7% frequency) and the most abundant baseline recipient variant (62.5% frequency). Furthermore, donor and recipient strains clustered together when compared to 10 control subjects from the same area and infected with the same HCV subtype. There was an increase in HVR1 complexity (number of genetic variants) over time in both spouses. In contrast, the 5'UTR region was stable and of low complexity throughout the study. In conclusion, intrafamilial HCV transmission may be established by a very minor variant and investigation of this phenomenon requires high-sensitivity assays, such as NGS.

Unbiased metagenomic sequencing complements specific routine diagnostic methods and increases chances to detect rare viral strains.

Multiplex PCR assays for respiratory viruses are widely used in routine diagnostics, as they are highly sensitive, rapid, and cost effective. However, depending on the assay system, cross-reactivity between viruses that share a high sequence homology as well as detection of rare virus isolates with sequence variations can be problematic. Virus sequence-independent metagenomic high-throughput sequencing allows for accurate detection of all virus species in a given sample, as we demonstrate here for human Enterovirus and Rhinovirus in a lung transplant patient. While early in infection a commercial PCR assay recorded Rhinovirus, high-throughput sequencing correctly identified human Enterovirus C104 as the source of infection, highlighting the potential of the technology and the benefit of applying open assay formats in complex diagnostic situations.

High-throughput sequencing of human immunoglobulin variable regions with subtype identification.

The humoral immune response plays a critical role in controlling infection, and the rapid adaptation to a broad range of pathogens depends on a highly diverse antibody repertoire. The advent of high-throughput sequencing technologies in the past decade has enabled insights into this immense diversity. However, not only the variable, but also the constant region of antibodies determines their in vivo activity. Antibody isotypes differ in effector functions and are thought to play a defining role in elicitation of immune responses, both in natural infection and in vaccination. We have developed an Illumina MiSeq high-throughput sequencing protocol that allows determination of the human IgG subtype alongside sequencing full-length antibody variable heavy chain regions. We thereby took advantage of the Illumina procedure containing two additional short reads as identifiers. By performing paired-end sequencing of the variable regions and customizing one of the identifier sequences to distinguish IgG subtypes, IgG transcripts with linked information of variable regions and IgG subtype can be retrieved. We applied our new method to the analysis of the IgG variable region repertoire from PBMC of an HIV-1 infected individual confirmed to have serum antibody reactivity to the Membrane Proximal External Region (MPER) of gp41. We found that IgG3 subtype frequencies in the memory B cell compartment increased after halted treatment and coincided with increased plasma antibody reactivity against the MPER domain. The sequencing strategy we developed is not restricted to analysis of IgG. It can be adopted for any Ig subtyping and beyond that for any research question where phasing of distant regions on the same amplicon is needed.

Prevalence and predictors for homo- and heterosubtypic antibodies against influenza a virus.

BACKGROUND: The effectiveness of trivalent influenza vaccination has been confirmed in several studies. To date, it is not known whether repeated exposure and vaccination to influenza promote production of cross-reactive antibodies. Furthermore, how strains encountered earlier in life imprint the immune response is currently poorly understood. METHODS: To determine the prevalence for human homo- and heterosubtypic antibody responses, we scrutinized serum samples from 305 healthy volunteers for hemagglutinin-binding and -neutralizing antibodies against several strains and subtypes of influenza A. Statistical analyses were then performed to establish the association of measured values with potential predictors. RESULTS: It was found that vaccination not only promoted higher binding and neutralizing antibody titers to homosubtypic influenza isolates but also increased heterosubtypic human immune responses. Both binding and neutralizing antibody titers in relation with age of the donors mirrored the course of the different influenza strain circulation during the last century. Advanced age appeared to be of advantage for both binding and neutralizing titers to most subtypes. In contrast, the first virus subtype encountered was found to imprint to some degree subsequent antibody responses. Antibodies to recent strains, however, primarily seemed to be promoted by vaccination. CONCLUSIONS: We provide evidence that vaccinations stimulate both homo- and heterosubtypic immune responses in young and middle-aged as well as more senior individuals. Our analyses suggest that influenza vaccinations not only prevent infection against currently circulating strains but can also stimulate broader humoral immune responses that potentially attenuate infections with zoonotic or antigenically shifted strains.

Deep sequencing of hepatitis C virus hypervariable region 1 reveals no correlation between genetic heterogeneity and antiviral treatment outcome.

BACKGROUND: Hypervariable region 1 (HVR1) contained within envelope protein 2 (E2) gene is the most variable part of HCV genome and its translation product is a major target for the host immune response. Variability within HVR1 may facilitate evasion of the immune response and could affect treatment outcome. The aim of the study was to analyze the impact of HVR1 heterogeneity employing sensitive ultra-deep sequencing, on the outcome of PEG-IFN-α (pegylated interferon α) and ribavirin treatment. METHODS: HVR1 sequences were amplified from pretreatment serum samples of 25 patients infected with genotype 1b HCV (12 responders and 13 non-responders) and were subjected to pyrosequencing (GS Junior, 454/Roche). Reads were corrected for sequencing error using ShoRAH software, while population reconstruction was done using three different minimal variant frequency cut-offs of 1%, 2% and 5%. Statistical analysis was done using Mann-Whitney and Fisher's exact tests. RESULTS: Complexity, Shannon entropy, nucleotide diversity per site, genetic distance and the number of genetic substitutions were not significantly different between responders and non-responders, when analyzing viral populations at any of the three frequencies (≥1%, ≥2% and ≥5%). When clonal sample was used to determine pyrosequencing error, 4% of reads were found to be incorrect and the most abundant variant was present at a frequency of 1.48%. Use of ShoRAH reduced the sequencing error to 1%, with the most abundant erroneous variant present at frequency of 0.5%. CONCLUSIONS: While deep sequencing revealed complex genetic heterogeneity of HVR1 in chronic hepatitis C patients, there was no correlation between treatment outcome and any of the analyzed quasispecies parameters.

HIV Haplotype Inference Using a Propagating Dirichlet Process Mixture Model.

This paper presents a new computational technique for the identification of HIV haplotypes. HIV tends to generate many potentially drug-resistant mutants within the HIV-infected patient and being able to identify these different mutants is important for efficient drug administration. With the view of identifying the mutants, we aim at analyzing short deep sequencing data called reads. From a statistical perspective, the analysis of such data can be regarded as a nonstandard clustering problem due to missing pairwise similarity measures between non-overlapping reads. To overcome this problem we propagate a Dirichlet Process Mixture Model by sequentially updating the prior information from successive local analyses. The model is verified using both simulated and real sequencing data.

Next-generation sequencing of HIV-1 RNA genomes: determination of error rates and minimizing artificial recombination.

Next-generation sequencing (NGS) is a valuable tool for the detection and quantification of HIV-1 variants in vivo. However, these technologies require detailed characterization and control of artificially induced errors to be applicable for accurate haplotype reconstruction. To investigate the occurrence of substitutions, insertions, and deletions at the individual steps of RT-PCR and NGS, 454 pyrosequencing was performed on amplified and non-amplified HIV-1 genomes. Artificial recombination was explored by mixing five different HIV-1 clonal strains (5-virus-mix) and applying different RT-PCR conditions followed by 454 pyrosequencing. Error rates ranged from 0.04-0.66% and were similar in amplified and non-amplified samples. Discrepancies were observed between forward and reverse reads, indicating that most errors were introduced during the pyrosequencing step. Using the 5-virus-mix, non-optimized, standard RT-PCR conditions introduced artificial recombinants in a fraction of at least 30% of the reads that subsequently led to an underestimation of true haplotype frequencies. We minimized the fraction of recombinants down to 0.9-2.6% by optimized, artifact-reducing RT-PCR conditions. This approach enabled correct haplotype reconstruction and frequency estimations consistent with reference data obtained by single genome amplification. RT-PCR conditions are crucial for correct frequency estimation and analysis of haplotypes in heterogeneous virus populations. We developed an RT-PCR procedure to generate NGS data useful for reliable haplotype reconstruction and quantification.

Accurate single nucleotide variant detection in viral populations by combining probabilistic clustering with a statistical test of strand bias.

BACKGROUND: Deep sequencing is a powerful tool for assessing viral genetic diversity. Such experiments harness the high coverage afforded by next generation sequencing protocols by treating sequencing reads as a population sample. Distinguishing true single nucleotide variants (SNVs) from sequencing errors remains challenging, however. Current protocols are characterised by high false positive rates, with results requiring time consuming manual checking. RESULTS: By statistical modelling, we show that if multiple variant sites are considered at once, SNVs can be called reliably from high coverage viral deep sequencing data at frequencies lower than the error rate of the sequencing technology, and that SNV calling accuracy increases as true sequence diversity within a read length increases. We demonstrate these findings on two control data sets, showing that SNV detection is more reliable on a high diversity human immunodeficiency virus sample as compared to a moderate diversity sample of hepatitis C virus. Finally, we show that in situations where probabilistic clustering retains false positive SNVs (for instance due to insufficient sample diversity or systematic errors), applying a strand bias test based on a beta-binomial model of forward read distribution can improve precision, with negligible cost to true positive recall. CONCLUSIONS: By combining probabilistic clustering (implemented in the program ShoRAH) with a statistical test of strand bias, SNVs may be called from deeply sequenced viral populations with high accuracy.