Characterizing and measuring bias in sequence data

Background

DNA sequencing technologies deviate from the ideal uniform distribution of reads. These biases impair scientific and medical applications. Accordingly, we have developed computational methods for discovering, describing and measuring bias.

Results

We applied these methods to the Illumina, Ion Torrent, Pacific Biosciences and Complete Genomics sequencing platforms, using data from human and from a set of microbes with diverse base compositions. As in previous work, library construction conditions significantly influence sequencing bias. Pacific Biosciences coverage levels are the least biased, followed by Illumina, although all technologies exhibit error-rate biases in high- and low-GC regions and at long homopolymer runs. The GC-rich regions prone to low coverage include a number of human promoters, so we therefore catalog 1,000 that were exceptionally resistant to sequencing. Our results indicate that combining data from two technologies can reduce coverage bias if the biases in the component technologies are complementary and of similar magnitude. Analysis of Illumina data representing 120-fold coverage of a well-studied human sample reveals that 0.20% of the autosomal genome was covered at less than 10% of the genome-wide average. Excluding locations that were similar to known bias motifs or likely due to sample-reference variations left only 0.045% of the autosomal genome with unexplained poor coverage.

Conclusions

The assays presented in this paper provide a comprehensive view of sequencing bias, which can be used to drive laboratory improvements and to monitor production processes. Development guided by these assays should result in improved genome assemblies and better coverage of biologically important loci.     

Analysis of the genetic diversity of influenza A viruses using next-generation DNA sequencing

Background

Influenza viruses exist as a large group of closely related viral genomes, also called quasispecies. The composition of this influenza viral quasispecies can be determined by an accurate and sensitive sequencing technique and data analysis pipeline. We compared the suitability of two benchtop next-generation sequencers for whole genome influenza A quasispecies analysis: the Illumina MiSeq sequencing-by-synthesis and the Ion Torrent PGM semiconductor sequencing technique.

Results

We first compared the accuracy and sensitivity of both sequencers using plasmid DNA and different ratios of wild type and mutant plasmid. Illumina MiSeq sequencing reads were one and a half times more accurate than those of the Ion Torrent PGM. The majority of sequencing errors were substitutions on the Illumina MiSeq and insertions and deletions, mostly in homopolymer regions, on the Ion Torrent PGM. To evaluate the suitability of the two techniques for determining the genome diversity of influenza A virus, we generated plasmid-derived PR8 virus and grew this virus in vitro. We also optimized an RT-PCR protocol to obtain uniform coverage of all eight genomic RNA segments. The sequencing reads obtained with both sequencers could successfully be assembled de novo into the segmented influenza virus genome. After mapping of the reads to the reference genome, we found that the detection limit for reliable recognition of variants in the viral genome required a frequency of 0.5% or higher. This threshold exceeds the background error rate resulting from the RT-PCR reaction and the sequencing method. Most of the variants in the PR8 virus genome were present in hemagglutinin, and these mutations were detected by both sequencers.

Conclusions

Our approach underlines the power and limitations of two commonly used next-generation sequencers for the analysis of influenza virus gene diversity. We conclude that the Illumina MiSeq platform is better suited for detecting variant sequences whereas the Ion Torrent PGM platform has a shorter turnaround time. The data analysis pipeline that we propose here will also help to standardize variant calling in small RNA genomes based on next-generation sequencing data.     

Cross-sectional detection of acute HIV infection: timing of transmission, inflammation and antiretroviral therapy

Background

Acute HIV infection (AHI) is a critical phase of infection when irreparable damage to the immune system occurs and subjects are very infectious. We studied subjects with AHI prospectively to develop better treatment and public health interventions.

Methods

Cross-sectional screening was employed to detect HIV RNA positive, antibody negative subjects. Date of HIV acquisition was estimated from clinical history and correlated with sequence diversity assessed by single genome amplification (SGA). Twenty-two cytokines/chemokines were measured from enrollment through week 24.

Results

Thirty-seven AHI subjects were studied. In 7 participants with limited exposure windows, the median exposure to HIV occurred 14 days before symptom onset. Lack of viral sequence diversification confirmed the short duration of infection. Transmission dates estimated by SGA/sequencing using molecular clock models correlated with transmission dates estimated by symptom onset in individuals infected with single HIV variants (mean of 28 versus 33 days). Only 10 of 22 cytokines/chemokines were significantly elevated among AHI participants at enrollment compared to uninfected controls, and only 4 participants remained seronegative at enrollment.

Discussion

The results emphasize the difficulty in recruiting subjects early in AHI. Viral sequence diversity proved accurate in estimating time of infection. Regardless of aggressive screening, peak viremia and inflammation occurred before enrollment and potential intervention. Given the personal and public health importance, improved AHI detection is urgently needed.     

Genetic heterogeneity of hepatitis C virus in association with antiviral therapy determined by ultra-deep sequencing

Background and Aims

The hepatitis C virus (HCV) invariably shows wide heterogeneity in infected patients, referred to as a quasispecies population. Massive amounts of genetic information due to the abundance of HCV variants could be an obstacle to evaluate the viral genetic heterogeneity in detail.

Methods

Using a newly developed massive-parallel ultra-deep sequencing technique, we investigated the viral genetic heterogeneity in 27 chronic hepatitis C patients receiving peg-interferon (IFN) α2b plus ribavirin therapy.

Results

Ultra-deep sequencing determined a total of more than 10 million nucleotides of the HCV genome, corresponding to a mean of more than 1000 clones in each specimen, and unveiled extremely high genetic heterogeneity in the genotype 1b HCV population. There was no significant difference in the level of viral complexity between immediate virologic responders and non-responders at baseline (p = 0.39). Immediate virologic responders (n = 8) showed a significant reduction in the genetic complexity spanning all the viral genetic regions at the early phase of IFN administration (p = 0.037). In contrast, non-virologic responders (n = 8) showed no significant changes in the level of viral quasispecies (p = 0.12), indicating that very few viral clones are sensitive to IFN treatment. We also demonstrated that clones resistant to direct-acting antivirals for HCV, such as viral protease and polymerase inhibitors, preexist with various abundances in all 27 treatment-naïve patients, suggesting the risk of the development of drug resistance against these agents.

Conclusion

Use of the ultra-deep sequencing technology revealed massive genetic heterogeneity of HCV, which has important implications regarding the treatment response and outcome of antiviral therapy.     

Distinguishing low frequency mutations from RT-PCR and sequence errors in viral deep sequencing data

Background

RNA viruses have high mutation rates and exist within their hosts as large, complex and heterogeneous populations, comprising a spectrum of related but non-identical genome sequences. Next generation sequencing is revolutionising the study of viral populations by enabling the ultra deep sequencing of their genomes, and the subsequent identification of the full spectrum of variants within the population. Identification of low frequency variants is important for our understanding of mutational dynamics, disease progression, immune pressure, and for the detection of drug resistant or pathogenic mutations. However, the current challenge is to accurately model the errors in the sequence data and distinguish real viral variants, particularly those that exist at low frequency, from errors introduced during sequencing and sample processing, which can both be substantial.

Results

We have created a novel set of laboratory control samples that are derived from a plasmid containing a full-length viral genome with extremely limited diversity in the starting population. One sample was sequenced without PCR amplification whilst the other samples were subjected to increasing amounts of RT and PCR amplification prior to ultra-deep sequencing. This enabled the level of error introduced by the RT and PCR processes to be assessed and minimum frequency thresholds to be set for true viral variant identification. We developed a genome-scale computational model of the sample processing and NGS calling process to gain a detailed understanding of the errors at each step, which predicted that RT and PCR errors are more likely to occur at some genomic sites than others. The model can also be used to investigate whether the number of observed mutations at a given site of interest is greater than would be expected from processing errors alone in any NGS data set. After providing basic sample processing information and the site’s coverage and quality scores, the model utilises the fitted RT-PCR error distributions to simulate the number of mutations that would be observed from processing errors alone.

Conclusions

These data sets and models provide an effective means of separating true viral mutations from those erroneously introduced during sample processing and sequencing.

Electronic supplementary material

The online version of this article (doi:10.1186/s12864-015-1456-x) contains supplementary material, which is available to authorized users.     

Genetic analysis of HIV-1 subtypes in Nairobi, Kenya

Background

Genetic analysis of a viral infection helps in following its spread in a given population, in tracking the routes of infection and, where applicable, in vaccine design. Additionally, sequence analysis of the viral genome provides information about patterns of genetic divergence that may have occurred during viral evolution.

Objective

In this study we have analyzed the subtypes of Human Immunodeficiency Virus -1 (HIV-1) circulating in a diverse sample population of Nairobi, Kenya.

Methodology

69 blood samples were collected from a diverse subject population attending the Aga Khan University Hospital in Nairobi, Kenya. Total DNA was extracted from peripheral blood mononuclear cells (PBMCs), and used in a Polymerase Chain Reaction (PCR) to amplify the HIV gag gene. The PCR amplimers were partially sequenced, and alignment and phylogenetic analysis of these sequences was performed using the Los Alamos HIV Database.

Results

Blood samples from 69 HIV-1 infected subjects from varying ethnic backgrounds were analyzed. Sequence alignment and phylogenetic analysis showed 39 isolates to be subtype A, 13 subtype D, 7 subtype C, 3 subtype AD and CRF01_AE, 2 subtype G and 1 subtype AC and 1 AG. Deeper phylogenetic analysis revealed HIV subtype A sequences to be highly divergent as compared to subtypes D and C.

Conclusion

Our analysis indicates that HIV-1 subtypes in the Nairobi province of Kenya are dominated by a genetically diverse clade A. Additionally, the prevalence of highly divergent, complex subtypes, intersubtypes, and the recombinant forms indicates viral mixing in Kenyan population, possibly as a result of dual infections.     

Whole genome sequencing of an ethnic Pathan (Pakhtun) from the north-west of Pakistan

Background

Pakistan covers a key geographic area in human history, being both part of the Indus River region that acted as one of the cradles of civilization and as a link between Western Eurasia and Eastern Asia. This region is inhabited by a number of distinct ethnic groups, the largest being the Punjabi, Pathan (Pakhtuns), Sindhi, and Baloch.

Results

We analyzed the first ethnic male Pathan genome by sequencing it to 29.7-fold coverage using the Illumina HiSeq2000 platform. A total of 3.8 million single nucleotide variations (SNVs) and 0.5 million small indels were identified by comparing with the human reference genome. Among the SNVs, 129,441 were novel, and 10,315 nonsynonymous SNVs were found in 5,344 genes. SNVs were annotated for health consequences and high risk diseases, as well as possible influences on drug efficacy. We confirmed that the Pathan genome presented here is representative of this ethnic group by comparing it to a panel of Central Asians from the HGDP-CEPH panels typed for ~650 k SNPs. The mtDNA (H2) and Y haplogroup (L1) of this individual were also typical of his geographic region of origin. Finally, we reconstruct the demographic history by PSMC, which highlights a recent increase in effective population size compatible with admixture between European and Asian lineages expected in this geographic region.

Conclusions

We present a whole-genome sequence and analyses of an ethnic Pathan from the north-west province of Pakistan. It is a useful resource to understand genetic variation and human migration across the whole Asian continent.

Electronic supplementary material

The online version of this article (doi:10.1186/s12864-015-1290-1) contains supplementary material, which is available to authorized users.     

Evaluation of viral genome assembly and diversity estimation in deep metagenomes

Background

Viruses have unique properties, small genome and regions of high similarity, whose effects on metagenomic assemblies have not been characterized so far. This study uses diverse in silico simulated viromes to evaluate how extensively genomes can be assembled using different sequencing platforms and assemblers. Further, it investigates the suitability of different methods to estimate viral diversity in metagenomes.

Results

We created in silico metagenomes mimicking various platforms at different sequencing depths. The CLC assembler revealed subpar compared to IDBA_UD and CAMERA , which are metagenomic-specific. Up to a saturation point, Illumina platforms proved more capable of reconstructing large portions of viral genomes compared to 454. Read length was an important factor for limiting chimericity, while scaffolding marginally improved contig length and accuracy. The genome length of the various viruses in the metagenomes did not significantly affect genome reconstruction, but the co-existence of highly similar genomes was detrimental. When evaluating diversity estimation tools, we found that PHACCS results were more accurate than those from CatchAll and clustering, which were both orders of magnitude above expected.

Conclusions

Assemblers designed specifically for the analysis of metagenomes should be used to facilitate the creation of high-quality long contigs. Despite the high coverage possible, scientists should not expect to always obtain complete genomes, because their reconstruction may be hindered by co-existing species bearing highly similar genomic regions. Further development of metagenomics-oriented assemblers may help bypass these limitations in future studies. Meanwhile, the lack of fully reconstructed communities keeps methods to estimate viral diversity relevant. While none of the three methods tested had absolute precision, only PHACCS was deemed suitable for comparative studies.

Electronic supplementary material

The online version of this article (doi:10.1186/1471-2164-15-989) contains supplementary material, which is available to authorized users.     

Whole Genome Amplification and De novo Assembly of Single Bacterial Cells

Background

Single-cell genome sequencing has the potential to allow the in-depth exploration of the vast genetic diversity found in uncultured microbes. We used the marine cyanobacterium Prochlorococcus as a model system for addressing important challenges facing high-throughput whole genome amplification (WGA) and complete genome sequencing of individual cells.

Methodology/Principal Findings

We describe a pipeline that enables single-cell WGA on hundreds of cells at a time while virtually eliminating non-target DNA from the reactions. We further developed a post-amplification normalization procedure that mitigates extreme variations in sequencing coverage associated with multiple displacement amplification (MDA), and demonstrated that the procedure increased sequencing efficiency and facilitated genome assembly. We report genome recovery as high as 99.6% with reference-guided assembly, and 95% with de novo assembly starting from a single cell. We also analyzed the impact of chimera formation during MDA on de novo assembly, and discuss strategies to minimize the presence of incorrectly joined regions in contigs.

Conclusions/Significance

The methods describe in this paper will be useful for sequencing genomes of individual cells from a variety of samples.     

An improved genome of the model marine alga Ostreococcus tauri unfolds by assessing Illumina de novo assemblies

Background

Cost effective next generation sequencing technologies now enable the production of genomic datasets for many novel planktonic eukaryotes, representing an understudied reservoir of genetic diversity. O. tauri is the smallest free-living photosynthetic eukaryote known to date, a coccoid green alga that was first isolated in 1995 in a lagoon by the Mediterranean sea. Its simple features, ease of culture and the sequencing of its 13 Mb haploid nuclear genome have promoted this microalga as a new model organism for cell biology. Here, we investigated the quality of genome assemblies of Illumina GAIIx 75 bp paired-end reads from Ostreococcus tauri, thereby also improving the existing assembly and showing the genome to be stably maintained in culture.

Results

The 3 assemblers used, ABySS, CLCBio and Velvet, produced 95% complete genomes in 1402 to 2080 scaffolds with a very low rate of misassembly. Reciprocally, these assemblies improved the original genome assembly by filling in 930 gaps. Combined with additional analysis of raw reads and PCR sequencing effort, 1194 gaps have been solved in total adding up to 460 kb of sequence. Mapping of RNAseq Illumina data on this updated genome led to a twofold reduction in the proportion of multi-exon protein coding genes, representing 19% of the total 7699 protein coding genes. The comparison of the DNA extracted in 2001 and 2009 revealed the fixation of 8 single nucleotide substitutions and 2 deletions during the approximately 6000 generations in the lab. The deletions either knocked out or truncated two predicted transmembrane proteins, including a glutamate-receptor like gene.

Conclusion

High coverage (>80 fold) paired-end Illumina sequencing enables a high quality 95% complete genome assembly of a compact ~13 Mb haploid eukaryote. This genome sequence has remained stable for 6000 generations of lab culture.

Electronic supplementary material

The online version of this article (doi:10.1186/1471-2164-15-1103) contains supplementary material, which is available to authorized users.     

Cost Effectiveness of Screening Strategies for Early Identification of HIV and HCV Infection in Injection Drug Users

Objective

To estimate the cost, effectiveness, and cost effectiveness of HIV and HCV screening of injection drug users (IDUs) in opioid replacement therapy (ORT).

Design

Dynamic compartmental model of HIV and HCV in a population of IDUs and non-IDUs for a representative U.S. urban center with 2.5 million adults (age 15–59).

Methods

We considered strategies of screening individuals in ORT for HIV, HCV, or both infections by antibody or antibody and viral RNA testing. We evaluated one-time and repeat screening at intervals from annually to once every 3 months. We calculated the number of HIV and HCV infections, quality-adjusted life years (QALYs), costs, and incremental cost-effectiveness ratios (ICERs).

Results

Adding HIV and HCV viral RNA testing to antibody testing averts 14.8–30.3 HIV and 3.7–7.7 HCV infections in a screened population of 26,100 IDUs entering ORT over 20 years, depending on screening frequency. Screening for HIV antibodies every 6 months costs $30,700/QALY gained. Screening for HIV antibodies and viral RNA every 6 months has an ICER of $65,900/QALY gained. Strategies including HCV testing have ICERs exceeding $100,000/QALY gained unless awareness of HCV-infection status results in a substantial reduction in needle-sharing behavior.

Discussion

Although annual screening for antibodies to HIV and HCV is modestly cost effective compared to no screening, more frequent screening for HIV provides additional benefit at less cost. Screening individuals in ORT every 3–6 months for HIV infection using both antibody and viral RNA technologies and initiating ART for acute HIV infection appears cost effective.     

Whole-genome analysis of Fusarium graminearum insertional mutants identifies virulence associated genes and unmasks untagged chromosomal deletions

Background

Identifying pathogen virulence genes required to cause disease is crucial to understand the mechanisms underlying the pathogenic process. Plasmid insertion mutagenesis of fungal protoplasts is frequently used for this purpose in filamentous ascomycetes. Post transformation, the mutant population is screened for loss of virulence to a specific plant or animal host. Identifying the insertion event has previously met with varying degrees of success, from a cleanly disrupted gene with minimal deletion of nucleotides at the insertion point to multiple-copy insertion events and large deletions of chromosomal regions. Currently, extensive mutant collections exist in laboratories globally where it was hitherto impossible to identify all the affected genes.

Results

We used a whole-genome sequencing (WGS) approach using Illumina HiSeq 2000 technology to investigate DNA tag insertion points and chromosomal deletion events in mutagenised, reduced virulence F. graminearum isolates identified in disease tests on wheat (Triticum aestivum). We developed the FindInsertSeq workflow to localise the DNA tag insertions to the nucleotide level. The workflow was tested using four mutants showing evidence of single and multi-copy insertions in DNA blot analysis. FindInsertSeq was able to identify both single and multi-copy concatenation insertion sites. By comparing sequencing coverage, unexpected molecular recombination events such as large tagged and untagged chromosomal deletions, and DNA amplification were observed in three of the analysed mutants. A random data sampling approach revealed the minimum genome coverage required to survey the F. graminearum genome for alterations.

Conclusions

This study demonstrates that whole-genome re-sequencing to 22x fold genome coverage is an efficient tool to characterise single and multi-copy insertion mutants in the filamentous ascomycete Fusarium graminearum. In some cases insertion events are accompanied with large untagged chromosomal deletions while in other cases a straight-forward insertion event could be confirmed. The FindInsertSeq analysis workflow presented in this study enables researchers to efficiently characterise insertion and deletion mutants.

Electronic supplementary material

The online version of this article (doi:10.1186/s12864-015-1412-9) contains supplementary material, which is available to authorized users.     

Use of a high resolution melting (HRM) assay to compare gag, pol, and env diversity in adults with different stages of HIV infection

Background

Cross-sectional assessment of HIV incidence relies on laboratory methods to discriminate between recent and non-recent HIV infection. Because HIV diversifies over time in infected individuals, HIV diversity may serve as a biomarker for assessing HIV incidence. We used a high resolution melting (HRM) diversity assay to compare HIV diversity in adults with different stages of HIV infection. This assay provides a single numeric HRM score that reflects the level of genetic diversity of HIV in a sample from an infected individual.

Methods

HIV diversity was measured in 203 adults: 20 with acute HIV infection (RNA positive, antibody negative), 116 with recent HIV infection (tested a median of 189 days after a previous negative HIV test, range 14–540 days), and 67 with non-recent HIV infection (HIV infected >2 years). HRM scores were generated for two regions in gag, one region in pol, and three regions in env.

Results

Median HRM scores were higher in non-recent infection than in recent infection for all six regions tested. In multivariate models, higher HRM scores in three of the six regions were independently associated with non-recent HIV infection.

Conclusions

The HRM diversity assay provides a simple, scalable method for measuring HIV diversity. HRM scores, which reflect the genetic diversity in a viral population, may be useful biomarkers for evaluation of HIV incidence, particularly if multiple regions of the HIV genome are examined.     

Rapid sequencing of the bamboo mitochondrial genome using Illumina technology and parallel episodic evolution of organelle genomes in grasses

Background

Compared to their counterparts in animals, the mitochondrial (mt) genomes of angiosperms exhibit a number of unique features. However, unravelling their evolution is hindered by the few completed genomes, of which are essentially Sanger sequenced. While next-generation sequencing technologies have revolutionized chloroplast genome sequencing, they are just beginning to be applied to angiosperm mt genomes. Chloroplast genomes of grasses (Poaceae) have undergone episodic evolution and the evolutionary rate was suggested to be correlated between chloroplast and mt genomes in Poaceae. It is interesting to investigate whether correlated rate change also occurred in grass mt genomes as expected under lineage effects. A time-calibrated phylogenetic tree is needed to examine rate change.

Methodology/Principal Findings

We determined a largely completed mt genome from a bamboo, Ferrocalamus rimosivaginus (Poaceae), through Illumina sequencing of total DNA. With combination of de novo and reference-guided assembly, 39.5-fold coverage Illumina reads were finally assembled into scaffolds totalling 432,839 bp. The assembled genome contains nearly the same genes as the completed mt genomes in Poaceae. For examining evolutionary rate in grass mt genomes, we reconstructed a phylogenetic tree including 22 taxa based on 31 mt genes. The topology of the well-resolved tree was almost identical to that inferred from chloroplast genome with only minor difference. The inconsistency possibly derived from long branch attraction in mtDNA tree. By calculating absolute substitution rates, we found significant rate change (∼4-fold) in mt genome before and after the diversification of Poaceae both in synonymous and nonsynonymous terms. Furthermore, the rate change was correlated with that of chloroplast genomes in grasses.

Conclusions/Significance

Our result demonstrates that it is a rapid and efficient approach to obtain angiosperm mt genome sequences using Illumina sequencing technology. The parallel episodic evolution of mt and chloroplast genomes in grasses is consistent with lineage effects.     

A universal protocol to generate consensus level genome sequences for foot-and-mouth disease virus and other positive-sense polyadenylated RNA viruses using the Illumina MiSeq

Background

Next-Generation Sequencing (NGS) is revolutionizing molecular epidemiology by providing new approaches to undertake whole genome sequencing (WGS) in diagnostic settings for a variety of human and veterinary pathogens. Previous sequencing protocols have been subject to biases such as those encountered during PCR amplification and cell culture, or are restricted by the need for large quantities of starting material. We describe here a simple and robust methodology for the generation of whole genome sequences on the Illumina MiSeq. This protocol is specific for foot-and-mouth disease virus (FMDV) or other polyadenylated RNA viruses and circumvents both the use of PCR and the requirement for large amounts of initial template.

Results

The protocol was successfully validated using five FMDV positive clinical samples from the 2001 epidemic in the United Kingdom, as well as a panel of representative viruses from all seven serotypes. In addition, this protocol was successfully used to recover 94% of an FMDV genome that had previously been identified as cell culture negative. Genome sequences from three other non-FMDV polyadenylated RNA viruses (EMCV, ERAV, VESV) were also obtained with minor protocol amendments. We calculated that a minimum coverage depth of 22 reads was required to produce an accurate consensus sequence for FMDV O. This was achieved in 5 FMDV/O/UKG isolates and the type O FMDV from the serotype panel with the exception of the 5′ genomic termini and area immediately flanking the poly(C) region.

Conclusions

We have developed a universal WGS method for FMDV and other polyadenylated RNA viruses. This method works successfully from a limited quantity of starting material and eliminates the requirement for genome-specific PCR amplification. This protocol has the potential to generate consensus-level sequences within a routine high-throughput diagnostic environment.

Electronic supplementary material

The online version of this article (doi:10.1186/1471-2164-15-828) contains supplementary material, which is available to authorized users.     

Factors influencing the emergence and spread of HIV drug resistance arising from rollout of antiretroviral pre-exposure prophylaxis (PrEP)

Background

The potential for emergence and spread of HIV drug resistance from rollout of antiretroviral (ARV) pre-exposure prophylaxis (PrEP) is an important public health concern. We investigated determinants of HIV drug resistance prevalence after PrEP implementation through mathematical modeling.

Methodology

A model incorporating heterogeneity in age, gender, sexual activity, HIV infection status, stage of disease, PrEP coverage/discontinuation, and HIV drug susceptibility, was designed to simulate the impact of PrEP on HIV prevention and drug resistance in a sub-Saharan epidemic.

Principal Findings

Analyses suggest that the prevalence of HIV drug resistance is influenced most by the extent and duration of inadvertent PrEP use in individuals already infected with HIV. Other key factors affecting drug resistance prevalence include the persistence time of transmitted resistance and the duration of inadvertent PrEP use in individuals who become infected on PrEP. From uncertainty analysis, the median overall prevalence of drug resistance at 10 years was predicted to be 9.2% (interquartile range 6.9%–12.2%). An optimistic scenario of 75% PrEP efficacy, 60% coverage of the susceptible population, and 5% inadvertent PrEP use predicts a rise in HIV drug resistance prevalence to only 2.5% after 10 years. By contrast, in a pessimistic scenario of 25% PrEP efficacy, 15% population coverage, and 25% inadvertent PrEP use, resistance prevalence increased to over 40%.

Conclusions

Inadvertent PrEP use in previously-infected individuals is the major determinant of HIV drug resistance prevalence arising from PrEP. Both the rate and duration of inadvertent PrEP use are key factors. PrEP rollout programs should include routine monitoring of HIV infection status to limit the spread of drug resistance.