Viral Metagenomics and Virome Sequencing: Detecting and Characterizing Viral Populations in Clinical and Environmental Samples

A single gram of human stool contains approximately 10⁹ virus-like particles. A liter of seawater holds roughly 10¹⁰. Phages outnumber bacteria ten to one in every ecosystem on Earth, driving horizontal gene transfer at a scale that reshapes microbial metabolism daily. Yet the vast majority of these viruses — often called the "viral dark matter" of microbiology — have never been cultured, sequenced, or classified. They do not carry 16S rRNA. They have no universal marker gene. For decades, they were invisible by design: every commonly used microbiome method, from amplicon sequencing to standard shotgun metagenomics, was built around bacterial targets.

Viral metagenomics — viromics — changes this. By enriching virus-like particles from a sample, extracting both DNA and RNA, and sequencing without the bacterial-size filter that standard metagenomics implicitly applies, viromics reveals the viral fraction of a microbiome at genomically meaningful depth. The technical gap between standard metagenomics and viromics is substantial, but a 2025 comparison of VLP-enriched versus bulk metagenomic methods showed that only roughly 27% of viral genomes overlap between the two approaches. If you only run standard shotgun metagenomics, you are missing approximately three-quarters of the virome.

This guide covers the end-to-end virome sequencing workflow: how to enrich viral particles from clinical and environmental samples, why certain amplification choices (particularly MDA) can distort community composition, how the bioinformatic toolchain for virus identification differs from the bacterial MAG pipeline, and where viromics is delivering actionable insight — from gut inflammatory disease to phage therapy screening to wastewater-based biosurveillance.

Figure 1: Virome sequencing workflow overview — from sample collection through viral enrichment, library preparation, sequencing, and bioinformatic analysis to functional insight.

Sample Preparation: Separating Viruses from Everything Else

The central challenge of viromics is that viruses are small, rare, and chemically fragile. A fecal sample is roughly 99% bacterial and host cells by biomass. A respiratory swab may contain picograms of viral nucleic acid in a background of micrograms of human RNA. Before any library is prepared, the viral fraction must be physically separated and concentrated.

• Viral Particle Enrichment

Three methods dominate, and the choice among them determines which viruses are captured.

Filtration is the universal first step. Passing a homogenized sample through sequential 0.8-μm and 0.45-μm or 0.22-μm polyethersulfone membrane filters removes bacteria, fungi, and cellular debris while allowing most viral particles to pass. For samples with high particulate load — soil, sediment, stool — pre-clarification by low-speed centrifugation is essential to prevent immediate filter clogging. Filtration alone yields a crude viral fraction suitable for some applications, but most protocols follow with a concentration step.

Ultracentrifugation at 75,000 to 100,000 × g for one to two hours pellets viral particles directly. It is the gold standard: it recovers both DNA and RNA viruses, works across sample types, and produces a clean viral fraction. The trade-off is equipment cost and throughput — a rotor holds perhaps six to twelve samples per run. For studies with hundreds of samples, this becomes a logistical bottleneck.

PEG precipitation uses polyethylene glycol 8000 at 10% (w/v) with 1 M sodium chloride, incubated overnight at 4°C, followed by centrifugation at roughly 1,200 × g. The pellet contains concentrated viral particles plus some co-precipitated bacterial debris. PEG precipitation costs a fraction of ultracentrifugation and scales to any batch size, making it the pragmatic choice for large cohort studies. A 2025 head-to-head comparison found that PEG precipitation and ultracentrifugation recovered broadly comparable viral communities from fecal samples, with PEG showing slightly higher yields and ultracentrifugation showing marginally higher purity.

In practice, the choice of enrichment method follows sample type and study scale: ultracentrifugation for high-purity viromics of up to twelve samples, PEG precipitation for large cohort studies where throughput and cost are primary, and filtration-only for resource-limited pilot studies. For water samples above one liter, tangential flow filtration (TFF) with a 100 kDa molecular weight cutoff membrane concentrates viral particles without a pelleting step, making it the preferred approach for seawater, wastewater, and freshwater viromics. TFF also concentrates viral particles while simultaneously removing dissolved PCR inhibitors such as humic acids, a dual benefit that simplifies downstream processing of complex environmental samples.

Regardless of the concentration method, a nuclease treatment step follows: DNase I plus RNase A (or a commercial benzonase cocktail) digests free nucleic acids released from lysed bacteria and host cells. This step is essential — without it, the "viral" fraction can contain up to 40% bacterial DNA, negating the enrichment effort. After nuclease treatment, the enzymes themselves must be inactivated or removed before viral capsids are lysed for nucleic acid extraction.

• Nucleic Acid Extraction: DNA and RNA Together

Most viruses of interest in a microbiome study have DNA genomes — bacteriophages, adenoviruses, anelloviruses — but RNA viruses including rotavirus, norovirus, and SARS-CoV-2 are critical in clinical and wastewater contexts. A single extraction that recovers both nucleic acid types, followed by separate DNA and RNA library preparation, is the most informative strategy.

Dedicated viral extraction kits — the QIAamp Viral RNA Mini Kit, the PureLink Viral RNA/DNA Mini Kit, and the AllPrep PowerViral DNA/RNA Kit — consistently outperform standard metagenomic extraction kits (designed for bacterial cells) for viral read enrichment. A 2025 comparison across multiple extraction methods confirmed that kits optimized for viral particle lysis and small-volume elution recover roughly 2- to 5-fold more viral reads per sample than bacterial-focused alternatives. Mechanical lysis is critical for recovery of structurally robust viral capsids. Bead-beating with 0.1-mm silica-zirconia beads for 3 to 5 minutes is standard practice for fecal and soil viromes. However, for enveloped RNA viruses — including SARS-CoV-2 and influenza — gentler chemical lysis with guanidinium thiocyanate-based buffers preserves the labile lipid envelope while still releasing the RNA genome. For RNA virus recovery specifically, carrier RNA should be omitted from the extraction — it competes during reverse transcription and suppresses viral signal.

CD Genomics' Viral Metagenomic Sequencing service performs viral particle enrichment and DNA/RNA co-extraction from a wide range of clinical and environmental sample types, with quality control at each step from filtration through nuclease treatment.

Figure 2: A three-panel comparison showing VLP enrichment methods — ultracentrifugation, PEG precipitation, and filtration-only — with yield and purity metrics for each approach.

Library Preparation: The Amplification Fork in the Road

Viral nucleic acid yields from VLP enrichment are often in the picogram-to-nanogram range — below the input threshold for standard ligation-based library preparation. Amplification is necessary, and the choice of amplification method determines whether the resulting library reflects the true viral community or an artifact of polymerase preference.

• The MDA Problem

Multiple displacement amplification, using φ29 DNA polymerase, was the default viromics amplification method for over a decade. φ29 amplifies circular single-stranded DNA with extreme efficiency — it is, after all, the polymerase bacteriophage use to replicate their own genomes in a rolling-circle mechanism.

The problem is what this does to community composition. MDA preferentially amplifies small circular single-stranded DNA through the same rolling-circle mechanism that φ29 uses for its own genome replication. A 2024 benchmark using synthetic viral mixtures confirmed that MDA treatment enriches circular ssDNA viruses — particularly Microviridae — by orders of magnitude relative to their true abundance, while an alternative T7 DNA polymerase approach preserved the original ssDNA-to-dsDNA ratio across both synthetic and complex fecal virome samples. For a study interested in the total viral community, MDA renders the data uninterpretable. The one defensible use case for MDA is when the specific research question concerns circular ssDNA viruses — Microviridae, Circoviridae, Geminiviridae — and nothing else.

• SISPA and Random Amplification

Sequence-independent single-primer amplification (SISPA), in which cDNA or DNA is tagged with a defined primer sequence and then PCR-amplified, introduces less compositional bias than MDA while achieving similar fold-amplification. A 2025 fecal virome protocol benchmark found that PCR-SISPA with 30 cycles recovered viral community structures that correlated well with unamplified libraries, whereas MDA-amplified libraries showed near-zero correlation with any other method.

For RNA viruses, reverse transcription with random hexamers or degenerate 9N primers followed by template-switching (the SMART approach) provides cDNA suitable for both short-read and long-read library preparation. The SMART-RNA-Metavirome platform, published in 2025, demonstrated that a template-switching RT step coupled with 9N degenerate priming achieves roughly 99.9% genome coverage for dengue virus at low titers, with the added benefit of inherent rRNA depletion — the template-switching mechanism does not capture capped rRNA.

• Library Construction and Sequencing Depth

For Illumina-based viromics, tagmentation-based library preparation (Illumina DNA Prep) from approximately 125 ng of amplified DNA is the current standard. Ligation-based protocols (TruSeq Nano) are an alternative for AT-rich viral genomes, where tagmentation can introduce coverage gaps.

Sequencing depth recommendations are shifting. Early virome studies often sequenced at 1 to 2 million read pairs per sample — sufficient to describe the dominant viral taxa but inadequate for rare or low-abundance viruses. A benchmark study involving 882 virome samples recommended 4 to 10 Gb per sample for comprehensive virome characterization, corresponding to roughly 13 to 33 million read pairs at 2×150 bp. At these depths, viral genomes present at 0.01% relative abundance become detectable. For large cohort studies where deep virome sequencing of every sample is cost-prohibitive, a staged strategy — shallow sequencing of all samples followed by deep sequencing of a subset — balances discovery power with budget.

For studies integrating viromics with standard metagenomics, CD Genomics' Metagenomic Shotgun Sequencing provides the bacterial and archaeal profile from the same sample, while Viral Metagenomic Sequencing recovers the viral fraction — together delivering the complementary view that the 2025 comparison studies demonstrated is essential for comprehensive microbiome characterization.

Figure 3: Three amplification strategies compared — MDA, PCR-SISPA, and SMART template-switching — showing community composition bias across each method using stacked bar charts of viral family abundance.

Bioinformatics for Viromics: A Distinct Toolchain

The bioinformatic pipeline for viromics diverges from the standard metagenomic pipeline at the classification step. Bacterial metagenomic classification uses k-mer matching against curated genome databases. Viral classification faces a fundamentally harder problem: most viruses in any given sample have no close relative in any reference database.

• Quality Control and Host Removal

The pre-processing steps parallel standard metagenomics: quality trimming with fastp and host genome removal with Bowtie 2 against the appropriate reference — hg38 for human clinical samples, the host plant genome for rhizosphere or phyllosphere viromics, or a combined database for environmental samples. For clinical samples, human read fractions of 50 to 90% are common, and failure to deplete them swamps the viral signal.

• Read-Based Classification

Kraken 2 with a viral-focused database provides rapid taxonomic assignment at the read level, but sensitivity is entirely database-dependent. A viral sequence from an uncharacterized phage family may be classified only to the level of "Caudoviricetes" — or not at all. DIAMOND BLASTX, aligning translated nucleotide reads against a protein database, is more sensitive for divergent viral sequences because protein sequence is more conserved than nucleotide sequence. The trade-off is speed: DIAMOND is approximately two orders of magnitude slower than Kraken 2 on the same dataset.

For clinical viromics, where the question is "Is pathogen X present?", read-based classification with tools like SeqScreen or Centrifuge provides a yes/no answer within hours of sequencing. For ecological viromics, where the question is "What is the structure and diversity of the viral community?", read-based classification is a first pass that must be followed by assembly.

• Assembly and Viral Contig Identification

MEGAHIT and metaSPAdes both assemble viral genomes from metagenomic data, but viral assembly presents unique challenges: low coverage, high variability, and the presence of integrated prophages within bacterial contigs. Co-assembly of reads from multiple related samples improves recovery of low-abundance viral genomes by pooling coverage.

Viral contig identification tools have matured rapidly. VirSorter2 uses a random forest classifier trained on viral hallmark genes and genomic features to distinguish viral from bacterial contigs. It is the most widely adopted tool and performs well for double-stranded DNA phages, which dominate most viromes. CheckV assesses viral genome completeness, identifies host contamination at contig ends, and estimates the quality tier of each viral sequence — providing quality metrics analogous to CheckM2 for bacterial MAGs. VIBRANT adds functional annotation, identifying auxiliary metabolic genes (AMGs) carried by phages — genes that phages use to redirect host metabolism during infection, such as photosystem genes in cyanophages or nucleotide metabolism genes in gut phages.

For a typical fecal virome, MEGAHIT assembly followed by VirSorter2 and CheckV identifies approximately 2,000 to 5,000 viral operational taxonomic units (vOTUs) per sample, of which 10 to 20% are complete or high-quality viral genomes.

Taxonomic classification of the validated viral contigs uses tools such as vConTACT2 or VPF-Class against the IMG/VR database, which now contains over 15 million viral sequences, or the newer geNomad framework that combines marker-gene and machine-learning approaches for simultaneous viral identification and taxonomic assignment. geNomad's integrated pipeline — viral identification, taxonomy, and host prediction in a single tool — reduces the computational overhead of running three separate tools sequentially and has been benchmarked to recover approximately 20% more viral contigs than VirSorter2 alone in soil and marine viromes, though the two tools are complementary and frequently used together.

• Prophage and CRISPR Analysis

Integrated prophages — bacteriophage genomes inserted into bacterial chromosomes — are invisible to VLP-based viromics because they are retained on the bacterial-size filter. Detecting them requires analysis of the bulk metagenomic assembly. Tools such as VIBRANT and geNomad flag prophage regions within bacterial contigs. CRISPR spacer-to-phage matching provides the strongest form of host-virus linkage: if a bacterial CRISPR array contains a spacer that matches a viral contig, that bacterium (or its ancestor) was infected by that phage. CRISPR arrays function as a bacterial adaptive immune memory — when a bacterium survives phage infection, it incorporates a short fragment of the phage genome into its own CRISPR locus. Matching these archived spacers to assembled viral contigs thus reveals which phages have historically infected which bacterial hosts, providing the strongest available evidence for host-virus linkage in metagenomic data. This host prediction information is essential for understanding phage-bacteria interaction networks but is only available when both bulk metagenomic and viromic data are analyzed together.

CD Genomics' virome analysis pipeline includes the full virome bioinformatic workflow: quality control, host read removal, Kraken 2 and DIAMOND-based classification, MEGAHIT assembly, VirSorter2 and CheckV viral contig validation, and taxonomic assignment against IMG/VR.

Figure 4: Virome bioinformatic pipeline — quality control → host removal → read-based classification (Kraken 2 / DIAMOND) → assembly (MEGAHIT) → viral contig identification (VirSorter2 / CheckV) → functional annotation (VIBRANT).

Applications: From Gut to Environment

The gut virome is dominated by bacteriophages — roughly 90 to 95% of viral reads in most fecal viromes map to Caudoviricetes (tailed phages) and Microviridae. The remaining fraction includes eukaryotic viruses (anelloviruses, adenoviruses, enteroviruses) and plant viruses from diet. Unlike the adult gut bacteriome, which stabilizes by roughly age three, the gut virome continues to shift through childhood and adolescence.

In inflammatory bowel disease, multiple independent cohorts have reported elevated Caudovirales richness in Crohn's disease and ulcerative colitis relative to healthy controls — the opposite of the bacterial diversity pattern, where disease is associated with reduced richness. A 2025 meta-analysis covering 2,066 metagenomic samples from 16 cohorts found that viral Shannon diversity was consistently higher in IBD across all studies, and that a random forest classifier trained on the top 50 discriminating vOTUs separated cases from controls with an area under the curve of 0.85 to 0.90.

Phage therapy — using lytic bacteriophages to treat multidrug-resistant bacterial infections — has driven renewed interest in environmental phage discovery. Metagenomic sequencing of wastewater, seawater, and soil samples identifies complete phage genomes carrying no known antibiotic resistance or toxin genes, which can then be screened against clinical bacterial isolates. The 882-sample virome benchmark dataset mentioned above was generated in part to support phage discovery for a Klebsiella pneumoniae phage therapy program.

In biosurveillance, wastewater viromics provides a population-level, pathogen-agnostic monitoring capability. During and after the COVID-19 pandemic, metagenomic sequencing of wastewater viral concentrates detected not only SARS-CoV-2 lineages but also enterovirus, norovirus, rotavirus, and hepatitis A virus — providing early warning of community transmission before clinical cases were reported. A 2024 review of metagenomic next-generation sequencing in infectious disease diagnosis documented the expanding role of this approach across respiratory, bloodstream, central nervous system, and gastrointestinal infections, confirming that mNGS detects clinically relevant viruses that culture and targeted PCR miss. The same approach applied to agricultural runoff and aquaculture effluent monitors the movement of viral pathogens across the animal-human interface.

A cross-reference to CD Genomics' Metagenomic Sequencing Services provides additional context on how viromics integrates with broader metagenomic strategies, including the gut microbiome workflows covered in our guide to Shotgun Metagenomic Sequencing for Gut Microbiome Studies.

Figure 5: Gut Virome Dysbiosis in IBD — elevated Caudoviricetes richness and host-virus network rewiring in inflammatory bowel disease, with violin plots of viral Shannon diversity and stacked bar charts of viral family relative abundance comparing healthy versus IBD cohorts.

Challenges and Limitations

The viral dark matter problem is the single largest analytical limitation. In the most recent benchmark study of environmental viromes, approximately 60 to 90% of assembled viral contigs had no significant match to any sequence in IMG/VR or RefSeq. These sequences represent novel viral lineages — entire families or orders without a single characterized representative. Protein structure-based classification using tools like AlphaFold-predicted structures shows early promise for placing these sequences on a viral tree of life: hallmark viral proteins such as major capsid proteins, terminases, and portal proteins retain structural folds over billions of years of divergence even when their primary sequences share no detectable similarity. However, predicting structures for thousands of novel viral contigs remains computationally demanding and is not yet routine in virome analysis pipelines.

Database completeness varies dramatically by environment. The human gut virome is the best-characterized, with tens of thousands of vOTUs cataloged across cohorts. Soil and marine viromes are far less complete, and the viromes of invertebrates, protists, and extreme environments remain largely uncharted. A viral sequence from a deep-sea hydrothermal vent has perhaps a 5 to 10% chance of matching anything in the database.

Quantification is the second unresolved challenge. Unlike 16S or standard metagenomics, where internal standard spike-ins now enable absolute abundance quantification in copies per gram or copies per cell, viral absolute quantification is complicated by the variable relationship between VLP count and genome copy number — a single infected bacterial cell can release hundreds of phage particles, each carrying one genome copy, in a lytic burst. For studies comparing viral abundance across conditions, relative abundance (reads per million or genome copies per million) remains standard, with the understanding that a relative increase in one viral taxon mathematically forces a decrease in others.

The cost of deep virome sequencing — 4 to 10 Gb per sample for comprehensive coverage — exceeds the cost of shallow shotgun metagenomics for bacteria. Studies with large cohorts must make deliberate trade-offs between sample number and per-sample depth. For biomarker discovery, deep virome sequencing of a discovery cohort followed by targeted validation (qPCR or digital PCR of the top candidate vOTUs) in a larger replication cohort is the practical strategy.

CD Genomics supports Viral Metagenomic Sequencing at flexible sequencing depths, from shallow virome surveys at 2 to 3 million reads per sample to deep virome characterization at 20 million-plus reads, with the option of Absolute Metagenomic Sequencing Service for studies requiring quantitative viral load estimates.

Figure 6: Viral dark matter visualization — a pie chart array showing the proportion of classified vs. unclassified viral contigs across human gut, soil, marine, and extreme environments, with representative 3D viral morphology icons for known families and question-mark silhouettes for predicted but unobserved viral shapes.

How CD Genomics Delivers Your Virome Sequencing Project

A virome sequencing project at CD Genomics follows a defined workflow optimized for viral recovery. Samples are received with chain-of-custody documentation and processed according to matrix-specific protocols: fecal samples are homogenized in DNA/RNA Shield, water samples are filtered through 0.22-μm membrane, and respiratory or tissue samples are homogenized in viral transport medium. Viral particle enrichment is performed by PEG precipitation or ultracentrifugation selected according to sample type and study goals, followed by nuclease treatment and extraction of total viral nucleic acid using kits optimized for viral particle lysis. Extracted nucleic acid is split into parallel DNA and RNA workflows — PCR-SISPA amplification and tagmentation for DNA viruses, reverse transcription and amplification for RNA viruses — and sequenced on the Illumina NovaSeq platform. The bioinformatic pipeline includes quality trimming, host read removal, Kraken 2 and DIAMOND BLASTX classification, MEGAHIT assembly, VirSorter2 viral contig identification, CheckV quality assessment, and taxonomic assignment against IMG/VR. Deliverables include raw FASTQ files, QC reports, viral taxonomic abundance tables at the species and vOTU level, assembled viral genomes with CheckV quality metrics, functional annotation of viral AMGs, and a comprehensive project report. Turnaround for a 50-sample virome project is approximately six to eight weeks from sample receipt.

For studies that require bacterial taxonomic and functional profiling alongside the viral analysis, CD Genomics' Metagenomic Shotgun Sequencing service provides the complementary bacterial view. For projects investigating actively replicating viruses through gene expression, Metatranscriptomic Sequencing captures the RNA fraction of both host and virus. For cohort-scale studies that combine viromics, metagenomics, and metabolomics, our Multi-Omics Service provides integrated multi-platform study design and data integration. When the virome survey identifies phage candidates for therapeutic development, Microbial Whole Genome Sequencing supports the bacterial host genome characterization needed to confirm phage-host specificity.

References:

1. De Chiara L, Romano V, Cerutti F, et al. A comparison of methods for the optimal recovery of the human fecal virome. medRxiv. 2025. doi:10.1101/2025.05.12.25327428 (CC BY 4.0)
2. Billaud M, Theodorou I, Lamy-Besnier Q, et al. T7 DNA polymerase treatment improves quantitative sequencing of both double-stranded and single-stranded DNA viruses. Peer Community Journal. 2024;4:e63. doi:10.24072/pcjournal.437 (CC BY 4.0)
3. Li N, Liu S, Zhang Y, et al. SMART-RNA-Metavirome: a practical RNA metavirome platform compatible with high-throughput sequencing of both short and long reads. Infectious Diseases of Poverty. 2025;14:101. doi:10.1186/s40249-025-01371-z (CC BY 4.0)
4. Nayfach S, Camargo AP, Schulz F, et al. CheckV assesses the quality and completeness of metagenome-assembled viral genomes. Nature Biotechnology. 2021;39:578-585. doi:10.1038/s41587-020-00774-7 (CC BY 4.0)
5. Guo J, Bolduc B, Zayed AA, et al. VirSorter2: a multi-classifier, expert-guided approach to detect diverse DNA and RNA viruses. Microbiome. 2021;9:37. doi:10.1186/s40168-020-00990-y (CC BY 4.0)
6. Ansari MH, Ebrahimi M, Fattahi MR, et al. Viral metagenomic analysis of fecal samples reveals an enteric virome signature in irritable bowel syndrome. BMC Microbiology. 2020;20:123. doi:10.1186/s12866-020-01817-4.
7. Moon K, Cho JC. Virome datasets and viral metagenome-assembled genomes from aquaculture-impacted freshwater environments. Scientific Data. 2026;13:284. doi:10.1038/s41597-026-07383-0 (CC BY 4.0)
8. Kieft K, Zhou Z, Anantharaman K. VIBRANT: automated recovery, annotation and curation of microbial viruses, and evaluation of viral community function from genomic sequences. Microbiome. 2020;8:90. doi:10.1186/s40168-020-00867-0 (CC BY 4.0)
9. Zhao Y, Zhang W, Wang X, et al. Application of metagenomic next-generation sequencing in the diagnosis of infectious diseases. Frontiers in Cellular and Infection Microbiology. 2024;14:1458316. doi:10.3389/fcimb.2024.1458316 (CC BY 4.0)

For research purposes only, not intended for clinical diagnosis, treatment, or individual health assessments.

For research purposes only, not intended for clinical diagnosis, treatment, or individual health assessments.