Enterovirus Genome Research Articles

Analysis of the Complete Genomes of Enterovirus 71 Subtypes in China.

Enterovirus 71 (EV-A71) is one of the most pathogens to hand, foot, and mouth disease (HFMD) as well as neurological complications in young children. Molecular characteristic of EV-A71 is important to prevent the virus outbreak. Here, the complete genomes of EV-A71 from China between 1998 and 2019 were downloaded from GenBank. The phylogenetic trees were developed by MEGA7.0 software, and the complete genetic epidemiological characteristics and amino acid mutations of EV-A71 from China were also analysed. The results showed that major epidemic EV-A71 subtype was C4b before 2004, while it turned to C4a after 2004 in mainland China, and C4 and B5 were major subtypes in Taiwan. VP1, VP4, 2C, 3C, 3D, and complete genome sequence can be used for virus genotyping, and VP1, VP4, and complete genomes have obvious advantages over other segments. There were many significant mutations in the viral complete genome sequence. This study indicated that the major C4 and B5 subtypes will contribute to the development of vaccines and drugs of EV-A71 for prevention and monitoring of EV-A71-associated HFMD in China.

Characterisation of enterovirus RNA detected in the pancreas and other specimens of live patients with newly diagnosed type 1 diabetes in the DiViD study

Aims/hypothesisThe Diabetes Virus Detection (DiViD) study is the first study to laparoscopically collect pancreatic tissue and purified pancreatic islets together with duodenal mucosa, serum, peripheral blood mononuclear cells (PBMCs) and stools from six live adult patients (age 24–35 years) with newly diagnosed type 1 diabetes. The presence of enterovirus (EV) in the pancreatic islets of these patients has previously been reported.MethodsIn the present study we used reverse transcription quantitative real-time PCR (RT-qPCR) and sequencing to characterise EV genomes present in different tissues to understand the nature of infection in these individuals.ResultsAll six patients were found to be EV-positive by RT-qPCR in at least one of the tested sample types. Four patients were EV-positive in purified islet culture medium, three in PBMCs, one in duodenal biopsy and two in stool, while serum was EV-negative in all individuals. Sequencing the 5′ untranslated region of these EVs suggested that all but one belonged to enterovirus B species. One patient was EV-positive in all these sample types except for serum. Sequence analysis revealed that the virus strain present in the isolated islets of this patient was different from the strain found in other sample types. None of the islet-resident viruses could be isolated using EV-permissive cell lines.Conclusions/interpretationEV RNA can be frequently detected in various tissues of patients with type 1 diabetes. At least in some patients, the EV strain in the pancreatic islets may represent a slowly replicating persisting virus.Graphical abstract

MicroRNAs miR-18a and miR-452 regulate the replication of enterovirus 71 by targeting the gene encoding VP3.

MicroRNAs (miRNAs) are crucial in the process of host-pathogen interaction. In this study, we established a screening system for miRNAs of target genes to detect the effect of miRNAs on Enterovirus 71 (EV71) replication in rhabdomyosarcoma (RD) cells. A 3'-untranslated region (UTR) dual-luciferase assay was performed to confirm putative miRNA targets in EV71 genome. Firstly, 13 fragments of EV71 genome were inserted into the vector pMIR, and luciferase activities were analyzed to identify the putative miRNAs of target genes. The expression of the reporter protein was significantly downregulated in cells transfected with the vector containing gene VP3. Then we screened for miRNAs that might target to VP3 through online analysis software. In addition, Western blot, real-time PCR, virus titration, and morphological changes were considered to examine the effects of miRNAs on virus replication. The results suggested that miR-18a and miR-452 repress the reproduction of EV71 virus by binding to VP3. Moreover, EV71 infection also affected the expression of endogenous miR-18a and miR-452. In addition, no significant cytotoxic effects were observed. The results from this study suggest that the intracellular miRNAs may play vital roles in the host-virus interaction.

ICAM-1 induced rearrangements of capsid and genome prime rhinovirus 14 for activation and uncoating

Most rhinoviruses, which are the leading cause of the common cold, utilize intercellular adhesion molecule-1 (ICAM-1) as a receptor to infect cells. To release their genomes, rhinoviruses convert to activated particles that contain pores in the capsid, lack minor capsid protein VP4, and have an altered genome organization. The binding of rhinoviruses to ICAM-1 promotes virus activation; however, the molecular details of the process remain unknown. Here, we present the structures of virion of rhinovirus 14 and its complex with ICAM-1 determined to resolutions of 2.6 and 2.4 Å, respectively. The cryo-electron microscopy reconstruction of rhinovirus 14 virions contains the resolved density of octanucleotide segments from the RNA genome that interact with VP2 subunits. We show that the binding of ICAM-1 to rhinovirus 14 is required to prime the virus for activation and genome release at acidic pH. Formation of the rhinovirus 14-ICAM-1 complex induces conformational changes to the rhinovirus 14 capsid, including translocation of the C termini of VP4 subunits, which become poised for release through pores that open in the capsids of activated particles. VP4 subunits with altered conformation block the RNA-VP2 interactions and expose patches of positively charged residues. The conformational changes to the capsid induce the redistribution of the virus genome by altering the capsid-RNA interactions. The restructuring of the rhinovirus 14 capsid and genome prepares the virions for conversion to activated particles. The high-resolution structure of rhinovirus 14 in complex with ICAM-1 explains how the binding of uncoating receptors enables enterovirus genome release.

The prevalence of enteroviral RNA and protein in mitral valves of chronic rheumatic heart disease.

Acute Rheumatic Fever/ Rheumatic Heart Disease (ARF/RHD), a sequel of group A streptococcal (GAS) infection, even today constitutes a public health issue in developing countries including India. Differences in the prevalence of ARF/RHD in countries with a similar prevalence of GAS infections indicate the role of other cofactors in pathogenesis of RHD. We investigated the prevalence of enterovirus (EV) in RHD by probing for both EV RNA and VP1 protein using Nonisotopic In Situ Hybridization (NISH) and Immunohistochemistry (IHC) respectively in 75 valvectomy specimens obtained from RHD cases. Twenty-eight (37%) of the valves showed tissue inflammation with lymphocytic infiltration in a majority of the cases. Twenty-six and 27 (38% and 40%) of the 68 valves showed the presence of EV by IHC and NISH respectively, indicating a very good association between the two tests; however, only about 46 to 48% of them exhibited tissue inflammation. In eight cases (12%) the EV genome was detectable in absence of VP1 protein perhaps indicating a latent viral infection. Due to a high degree of endemicity of EV in India, we are tempted to speculate that EV may be responsible for the severity and rapid progression of RHD. The virus could either be working synergistically with GAS or could be an opportunist infecting damaged valves.

Immune Transcriptome of Cells Infected with Enterovirus Strains Obtained from Cases of Autoimmune Thyroid Disease

Background: Hashimoto’s thyroiditis and Graves’ disease are autoimmune thyroid disorders (AITD) of unknown origin. Enterovirus (EV) infection of thyroid cells has been implicated as a possible initiator of cell damage and of organ-specific autoimmunity. We asked whether persistent infection of human epithelial cells with EV strains obtained from thyroid tissue of AITD patients could be associated with transcriptional changes capable of fostering immunopathology. Methods: EV isolates obtained from thyroid tissue of AITD cases were used to infect the AV3 epithelial cell line. AV3 cells incubated with a virus-free medium from thyroid tissue of subjects without evidence of thyroid autoimmunity were used as uninfected controls. Transcripts of immune-related genes were compared in infected vs. uninfected cells. Results: The EV genome and antigens were detected only in the cells exposed to AITD-derived virus isolates, not in control cells. Persistent EV infection, while suppressing transcription of several type I IFN and cytokine determinants, was associated with enhanced transcription of NFKB1/RELA, IFNAR1, JAK1/STAT1, i.e., the determinants that play key immunologic roles. Infection also led to upregulation of the CCL2 chemokine and the IL-18 pro-inflammatory interleukin. Conclusion: As in the case of EV strains obtained from autoimmune diabetes, results show that the EV strains that are present in the thyroid of AITD cases do repress IFN and cytokine pathways. JAK1/STAT1 upregulation supports activation of TLR pathways and aberrant T cell signaling. In the early phases of AITD, our results highlight the potential benefit of interventions aimed at blocking the viral infection and easing the inflammatory response.

Two Natural Recombination gave rise to the Coxsackievirus B3 GV that triggered outbreaks in China in 2006 - 2012

Coxsackievirus B3 serotype GV caused the epidemic of Coxsackievirus B3 infection in China from 2006 to 2012. To study the evolution and recombination of Coxsackievirus B3 serotype GV, we performed recombination and phylogenetic analysis of 499 complete genomes of Enterovirus B available in GenBank, dated April 2019. Results indicated that most of the strains of Coxsackievirus B3 GV in P1 region were derived from a Coxsackievirus B3 GVI parent, and in P2-3 region from EchoV E25 strain, with nucleotide identities of 97.2% and 94.7%, respectively. Other strains of Coxsackievirus B3 GV-C1 in P1-P2 regions were derived from Coxsackievirus B3 GV-C3, whereas those in P3 regions were from CVB5. These naturally occurring recombination events were confirmed by phylogenetic analysis. This study indicates that two naturally occurring recombination gave rise to the coxsackievirus B3 GV that triggered outbreaks in China in 2006 - 2012.

Construction and characterization of an infectious cDNA clone of enterovirus 71: a rapid method for rescuing infectious virus based on stable cells expressing T7 polymerase.

Enterovirus 71 (EV71) is a causative agent of hand, foot and, mouth disease (HFMD) in young children. It is valuable for virologists to develop a fast method to rescue infectious virus from a viral cDNA clone. Here, we report a method for rapid rescue of infectious EV71 by using cells expressing T7 polymerase. The full-length EV71 genome was amplified in one step by long-distance PCR with a T7 promoter at the 5' end, and the T7 polymerase gene was cloned into a lentivirus vector for construction of a stable cell line expressing T7 polymerase. The infectious virus was rapidly and efficiently rescued by single transfection of cells with the infectious cDNA clone. Further experiments showed that the rescued virus had characteristics similar to those of the parental virus. This method circumvented the difficulty in performing in vitro transcription of a long linear DNA to obtain high-quality RNA. The construction of the viral cDNA clone and the fast rescue of the infectious virus will greatly benefit future investigations.

Assembly of infectious enteroviruses depends on multiple, conserved genomic RNA-coat protein contacts.

Picornaviruses are important viral pathogens, but despite extensive study, the assembly process of their infectious virions is still incompletely understood, preventing the development of anti-viral strategies targeting this essential part of the life cycle. We report the identification, via RNA SELEX and bioinformatics, of multiple RNA sites across the genome of a typical enterovirus, enterovirus-E (EV-E), that each have affinity for the cognate viral capsid protein (CP) capsomer. Many of these sites are evolutionarily conserved across known EV-E variants, suggesting they play essential functional roles. Cryo-electron microscopy was used to reconstruct the EV-E particle at ~2.2 Å resolution, revealing extensive density for the genomic RNA. Relaxing the imposed symmetry within the reconstructed particles reveals multiple RNA-CP contacts, a first for any picornavirus. Conservative mutagenesis of the individual RNA-contacting amino acid side chains in EV-E, many of which are conserved across the enterovirus family including poliovirus, is lethal but does not interfere with replication or translation. Anti-EV-E and anti-poliovirus aptamers share sequence similarities with sites distributed across the poliovirus genome. These data are consistent with the hypothesis that these RNA-CP contacts are RNA Packaging Signals (PSs) that play vital roles in assembly and suggest that the RNA PSs are evolutionarily conserved between pathogens within the family, augmenting the current protein-only assembly paradigm for this family of viruses.

Enterovirus subtyping in a routine UK laboratory setting between 2013 and 2017.

Human enteroviruses (EV) are the leading cause of viral meningitis. EV genotyping is predominantly performed through amplification and sequencing of viral capsid protein-1 (VP1), frequently by national reference laboratories (NRLs). To determine the frequency of genotyping failure in our NRL-submitted samples and apply a superior alternative assay to resolve untyped specimens. We initially audited genotyping data received for a cohort of patients in the East Midlands, UK by the NRL between 2013 and 2017, then identified an alternative RT-PCR typing method by literature review and evaluated primers from both assays in silico against comprehensive publicly available genomic data. The alternative assay was further optimised and applied to archived nucleic acids from previously untypable samples. Genotyping data showed a significant increase in untypable EV strains through the study period (p = 0.0073). Typing failure appeared unrelated to sample type or viral load. In silico analyses of 2,201 EV genomes showed high levels of mismatch between reference assay primers and clinically significant EV-species, in contrast to a selected alternative semi-nested RT-PCR VP1-typing assay. This alternative assay, with minor modifications, successfully genotyped 23 of 24 previously untypable yet viable archived specimens (EV-A, n = 4; EV-B, n = 19). Phylogenetic analyses identified no predominant strain within NRL untypable isolates, suggesting sub-optimal reference assay sensitivity across EV species, in agreement with in silico analyses. This modified highly sensitive RT-PCR assay presents a suitable alternative to the current English national reference VP1-typing assay and is recommended in other settings experiencing typing failure.

WarmStart colorimetric RT-LAMP for the rapid, sensitive and specific detection of Enteroviruses A-D targeting the 5'UTR region.

The aim of the present study was to develop a colorimetric LAMP assay for the detection of enteroviruses belonging to species A-D targeting the 5' untranslated region (5' UTR) of enteroviruses genome. The RNA was converted to cDNA by the reverse transcriptase and then amplified via LAMP by the WarmStart®Bst DNA polymerase, simultaneously in a single reaction tube, so we shortened the reaction time to 50min. The sensitivity of the assay regarding Enterovirus B, C and D was determined to be 0·30 CCID50 assay-1 while the sensitivity for Enterovirus A was 3·00 CCID50 assay-1 . The assay demonstrated high specificity and sensitivity for the detection of 45 reference strains of Enteroviruses A-D and validated on 20 clinical isolates. This assay can be used as a diagnostic tool for the rapid, sensitive and specific detection of enteroviruses, easily implemented in small clinical and research laboratories since LAMP amplicons were visualized by colour changes eliminating the requirement for post-amplification processing steps. We developed a colorimetric assay ideal for field situations for the detection of enteroviruses, by targeting the 5' UTR. This assay demonstrated high specificity and sensitivity, based on its performance on 45 EV A-D reference strains, on 20 EV B clinical isolates and on three non-enteroviral RNA viruses.

Inhibitory antibodies identify unique sites of therapeutic vulnerability in rhinovirus and other enteroviruses

The existence of multiple serotypes renders vaccine development challenging for most viruses in the Enterovirus genus. An alternative and potentially more viable strategy for control of these viruses is to develop broad-spectrum antivirals by targeting highly conserved proteins that are indispensable for the virus life cycle, such as the 3C protease. Previously, two single-chain antibody fragments, YDF and GGVV, were reported to effectively inhibit human rhinovirus 14 proliferation. Here, we found that both single-chain antibody fragments target sites on the 3C protease that are distinct from its known drug site (peptidase active site) and possess different mechanisms of inhibition. YDF does not block the active site but instead noncompetitively inhibits 3C peptidase activity through an allosteric effect that is rarely seen for antibody protease inhibitors. Meanwhile, GGVV antagonizes the less-explored regulatory function of 3C in genome replication. The interaction between 3C and the viral genome 5' noncoding region has been reported to be important for enterovirus genome replication. Here, the interface between human rhinovirus 14 3C and its 5' noncoding region was probed by hydrogen-deuterium exchange coupled mass spectrometry and found to partially overlap with the interface between GGVV and 3C. Consistently, prebinding of GGVV completely abolishes interaction between human rhinovirus 14 3C and its 5' noncoding region. The epitopes of YDF and GGVV, therefore, represent two additional sites of therapeutic vulnerability in rhinovirus. Importantly, the GGVV epitope appears to be conserved across many enteroviruses, suggesting that it is a promising target for pan-enterovirus inhibitor screening and design.

Molecular characterization of enteroviruses among hospitalized patients in Greece, 2013–2015

BackgroundThere are only sporadic data for the circulation of Enteroviruses (EVs) in Greece with previous studies reporting mainly the presence of Echoviruses (E) and Coxsackie viruses (CV) B. ObjectivesWe carried out a surveillance study for the molecular characterization of EVs detected in hospitalized patients throughout Greece as well as a phylogenetic analysis of the most frequently encountered serotypes. Study designStools, cerebrospinal fluids, throat swabs and blood samples were collected from hospitalized patients with suspicion of EV infection. All samples were tested for EVs by rRT-PCR targeting the 5′ untranslated region of EV genome. For positive samples, PCR amplification and sequencing targeting a part of VP1 region was performed. ResultsWe examined 831 samples and 209 were positive for EVs with Enterovirus B species being the most frequently amplified. E30, CVB5 and E9 were the most frequent serotypes of Enterovirus B species, whereas CVA6 and EV-A71 the most frequent serotypes of Enterovirus A species. Evs were significantly detected more frequently in stool samples compared to other types of specimens. Phylogenetic analysis revealed that most EV-A71 strains clustered in the subgenogroups C2 whereas all the CVA6 strains belonged to sub-genotype D3. Additionally, two different lineages of E30 and three different clusters of E9 viruses circulated simultaneously in Greece. Our data indicated that most EV strains from Greece were similar to strains circulating throughout Europe during the same period. ConclusionsWe provide a comprehensive picture of EVs circulating in Greece which can be helpful to interpret trends in EV diseases by associating them with circulating serotypes.

Heparan sulfate attachment receptor is a major selection factor for attenuated enterovirus 71 mutants during cell culture adaptation.

Enterovirus 71 (EV71) is a causative agent of hand, foot, and mouth disease (HFMD). However, this infection is sometimes associated with severe neurological complications. Identification of neurovirulence determinants is important to understand the pathogenesis of EV71. One of the problems in evaluating EV71 virulence is that its genome sequence changes rapidly during replication in cultured cells. The factors that induce rapid mutations in the EV71 genome in cultured cells are unclear. Here, we illustrate the population dynamics during adaptation to RD-A cells using EV71 strains isolated from HFMD patients. We identified a reproducible amino acid substitution from glutamic acid (E) to glycine (G) or glutamine (Q) in residue 145 of the VP1 protein (VP1-145) after adaptation to RD-A cells, which was associated with attenuation in human scavenger receptor B2 transgenic (hSCARB2 tg) mice. Because previous reports demonstrated that VP1-145G and Q mutants efficiently infect cultured cells by binding to heparan sulfate (HS), we hypothesized that HS expressed on the cell surface is a major factor for this selection. Supporting this hypothesis, selection of the VP1-145 mutant was prevented by depletion of HS and overexpression of hSCARB2 in RD-A cells. In addition, this mutation promotes the acquisition of secondary amino acid substitutions at various positions of the EV71 capsid to increase its fitness in cultured cells. These results indicate that attachment receptors, especially HS, are important factors for selection of VP1-145 mutants and subsequent capsid mutations. Moreover, we offer an efficient method for isolation and propagation of EV71 virulent strains with minimal selection pressure for attenuation.

Virome diversity analysis reveals novel enteroviruses and a human picobirnavirus in stool samples from African green monkeys with diarrhea

It is important to identify viruses in animals because most infectious diseases in humans are caused by viruses of zoonotic origin. African green monkey is a widely used non-human primate model in biomedical investigations. In this study, total RNAs were extracted from stool samples of 10 African green monkeys with diarrhea. High-throughput sequencing was used to characterize viromes. PCR and Sanger sequencing were used to determine the full genome sequences. Great viral diversity was observed. The dominant viruses were enteroviruses and picobirnaviruses. Six enterovirus genomes and a picobirnavirus RNA-dependent RNA polymerase sequence were characterized. Five enteroviruses belonged to two putative new genotypes of species Enterovirus J. One enterovirus belonged to EV-A92. The picobirnavirus RNA-dependent RNA polymerase sequence had the highest nucleotide similarity (93.48%) with human picobirnavirus isolate GPBV6C2. The present study helped to identify the potential zoonotic viruses in African green monkeys. Further investigations are required to elucidate their pathogenic roles in animals and humans.

First genomic characterization of a Belgian Enterovirus C104 using sequence-independent Nanopore sequencing

Because of the enormous variation in their genome sequence, genotyping enteroviruses by standard methods can prove to be quite challenging. Nanopore sequencing offers the potential to overcome the limitations of older techniques, but thus far, only amplicon-based strategies have been used to sequence complete enterovirus genomes. By combining a sequence-independent, single primer amplification (SISPA) for cDNA generation with next-generation sequencing using the Oxford Nanopore MinION, complete enterovirus genomes can be obtained in an easy-to-use, sequence-independent manner. To demonstrate its usability, we applied this technique to determine the complete genome sequence of an enterovirus C104 strain, representing the first documented occurrence of this uncommon enterovirus strain in Belgium.

Next generation sequencing of human enterovirus strains from an outbreak of enterovirus A71 shows applicability to outbreak investigations

BackgroundThe most recent documented Australian outbreak of enterovirus A71 (EV-A71) occurred in Sydney from 2012 to 2013. Over a four-month period more than 100 children presented to four paediatric hospitals with encephalitic presentations including fever and myoclonic jerks. The heterogeneous presentations included typical encephalomyelitis, and cardiopulmonary complications. ObjectivesTo characterise the genomes of enterovirus strains circulating during the 2013 Sydney EV-A71 outbreak and determine their phylogeny, phylogeography and association between genome and clinical phenotype. Study designWe performed an analysis of enterovirus (EV) positive specimens from children presenting to hospitals in the greater Sydney region of Australia during the 2013 outbreak. We amplified near full-length genomes of EV, and used next generation sequencing technology to sequence the virus. We used phylogenetic/phylogeographic analysis to characterize the outbreak viruses. ResultsWe amplified and sequenced 23/63 (37 %) genomes, and identified the majority (61 %) as EV-A71. The EV-A71 sequences showed high level sequence homology to C4a genogroups of EV-A71 circulating in China and Vietnam during 2012-13. Phylogenetic analysis showed EV-A71 strains associated with more severe symptoms, including encephalitis or cardiopulmonary failure, grouped together more closely than those from patients with hand, foot and mouth disease. Amongst the non-EV-A71 sequences were five other EV subtypes (representing enterovirus subtypes A and B), reflecting the diversity of EV co-circulation within the community. ConclusionsThis is the first Australian study investigating the near full-length genome of EV strains identified during a known outbreak of EV-A71. EV-A71 sequences were very similar to strains circulating in Asia during the same time period. Whole genome sequencing offers additional information over routine diagnostic testing such as characterisation of emerging recombinant strains and inform vaccine design.

Recombination in Enteroviruses, a Multi-Step Modular Evolutionary Process.

RNA recombination is a major driving force in the evolution and genetic architecture shaping of enteroviruses. In particular, intertypic recombination is implicated in the emergence of most pathogenic circulating vaccine-derived polioviruses, which have caused numerous outbreaks of paralytic poliomyelitis worldwide. Recent experimental studies that relied on recombination cellular systems mimicking natural genetic exchanges between enteroviruses provided new insights into the molecular mechanisms of enterovirus recombination and enabled to define a new model of genetic plasticity for enteroviruses. Homologous intertypic recombinant enteroviruses that were observed in nature would be the final products of a multi-step process, during which precursor nonhomologous recombinant genomes are generated through an initial inter-genomic RNA recombination event and can then evolve into a diversity of fitter homologous recombinant genomes over subsequent intra-genomic rearrangements. Moreover, these experimental studies demonstrated that the enterovirus genome could be defined as a combination of genomic modules that can be preferentially exchanged through recombination, and enabled defining the boundaries of these recombination modules. These results provided the first experimental evidence supporting the theoretical model of enterovirus modular evolution previously elaborated from phylogenetic studies of circulating enterovirus strains. This review summarizes our current knowledge regarding the mechanisms of recombination in enteroviruses and presents a new evolutionary process that may apply to other RNA viruses.

Characterization of Norovirus and Other Human Enteric Viruses in Sewage and Stool Samples Through Next-Generation Sequencing

This study aimed to optimize a method to identify human enteric viruses in sewage and stool samples using random primed next-generation sequencing. We tested three methods, two employed virus enrichment based on the binding properties of the viral capsid using pig-mucin capture or by selecting viral RNA prior to library preparation through a capture using the SureSelect target enrichment. The third method was based on a non-specific biophysical precipitation with polyethylene glycol. Full genomes of a number of common human enteric viruses including norovirus, rotavirus, husavirus, enterovirus and astrovirus were obtained. In stool samples full norovirus genome were detected as well as partial enterovirus genome. A variety of norovirus sequences was detected in sewage samples, with genogroup II being more prevalent. Interestingly, the pig-mucin capture enhanced not only the recovery of norovirus and rotavirus but also recovery of astrovirus, sapovirus and husavirus. Documenting sewage virome using these methods provides information for molecular epidemiology and may be useful in developing strategies to prevent further spread of viruses.

A64 Viral sequence classification using deep learning algorithms

Sewage samples have a high potential benefit for surveillance of circulating pathogens because they are easy to obtain and reflect population-wide circulation of pathogens. These type of samples typically contain a great diversity of viruses. Therefore, one of the main challenges of metagenomic sequencing of sewage for surveillance is sequence annotation and interpretation. Especially for high-threat viruses, false positive signals can trigger unnecessary alerts, but true positives should not be missed. Annotation thus requires high sensitivity and specificity. To better interpret annotated reads for high-threat viruses, we attempt to determine how classifiable they are in a background of reads of closely related low-threat viruses. As an example, we attempted to distinguish poliovirus reads, a virus of high public health importance, from other enterovirus reads. A sequence-based deep learning algorithm was used to classify reads as either polio or non-polio enterovirus. Short reads were generated from 500 polio and 2,000 non-polio enterovirus genomes as a training set. By training the algorithm on this dataset we try to determine, on a single read level, which short reads can reliably be labeled as poliovirus and which cannot. After training the deep learning algorithm on the generated reads we were able to calculate the probability with which a read can be assigned to a poliovirus genome or a non-poliovirus genome. We show that the algorithm succeeds in classifying the reads with high accuracy. The probability of assigning the read to the correct class was related to the location in the genome to which the read mapped, which conformed with our expectations since some regions of the genome are more conserved than others. Classifying short reads of high-threat viral pathogens seems to be a promising application of sequence-based deep learning algorithms. Also, recent developments in software and hardware have facilitated the development and training of deep learning algorithms. Further plans of this work are to characterize the hard-to-classify regions of the poliovirus genome, build larger training databases, and expand on the current approach to other viruses.