# Viral Genomic Surveillance: Building an Analysis and Reporting Workflow


## Key Takeaways

- A robust viral genomic surveillance workflow necessitates meticulous documentation and quality control at every stage, from specimen collection (e.g., nasopharyngeal swabs for respiratory viruses, lesion swabs for mpox) to public health communication, as any weakness can compromise downstream conclusions.
- Metadata is critical for interpreting genomic data; standardized fields such as collection date (ISO 8601), geographic location, host species, and specimen type, alongside controlled vocabularies, are essential for programmatic analysis and avoiding misinterpretation of phylogenetic clusters.
- Quality assurance includes implementing positive and negative controls during nucleic acid extraction and library preparation, and post-sequencing assessment of read quality metrics, with a minimum of 10x coverage per base often required for reliable consensus genome generation.
- Phylogenetic inference, while valuable for understanding genetic relatedness, must be combined with epidemiological data and temporal metadata to avoid overstating transmission events, as phylogenetic trees depict shared ancestry, not direct transmission routes.
- Public health reporting requires translating genomic findings into actionable insights, including standardized lineage classification (e.g., Pango for SARS-CoV-2), annotation of key amino acid changes, and transparent reporting of limitations such as sampling bias and sequencing error rates.
- Common pitfalls in viral genomic surveillance include ignoring sampling bias, failing to track tool versions, and misinterpreting phylogenetic clusters; these can be mitigated by stratified sampling, using environment managers like Conda, and integrating epidemiological data.

---

This guide connects sampling, sequencing, metadata curation, phylogenetic analysis, quality assurance, and public health communication into a single reproducible workflow. It is written for public health laboratory scientists, epidemiologists, bioinformaticians, and anyone establishing or refining a viral surveillance pipeline. You will learn how to move from raw clinical specimens to actionable reports while managing uncertainty and maintaining rigor. The approach described here draws on established technical references including the [NCBI Bookshelf](https://www.ncbi.nlm.nih.gov/books/) for foundational protocols and the [EMBL-EBI Training](https://www.ebi.ac.uk/training/) resources for data handling best practices.

Viral genomic surveillance is not a single step. It is a chain of dependent actions where a weak link can invalidate downstream conclusions. Every stage, from sample collection to report dissemination, must be documented and quality controlled. The [Galaxy Training Network](https://training.galaxyproject.org/) offers workflow templates that illustrate how to connect these stages, but each surveillance program must adapt general principles to its specific pathogen, resources, and public health questions.

## At a Glance

| Step | Key Considerations | Tools or Standards |
|------|-------------------|-------------------|
| Sample collection | Appropriate specimen type, storage, transport | Standard operating procedures, clinical guidelines |
| Nucleic acid extraction | Inhibitor removal, yield, purity | Commercial kits with internal controls |
| Sequencing library prep | Target enrichment vs metagenomic, depth requirements | Amplicon panels, capture probes, or random priming |
| Sequencing | Platform choice, read length, error profile | Illumina, Nanopore, PacBio, run QC metrics |
| Data preprocessing | Adapter trimming, read quality filtering, host read removal | fastp, Trimmomatic, Kraken2 |
| Consensus generation | Reference mapping vs de novo assembly, minimum depth thresholds | iVar, bcftools, SPAdes |
| Metadata curation | Standardized fields, ontology terms, timestamps | NCBI BioSample attributes, GISAID metadata |
| Quality assurance | Contamination checks, phylogenetic placement consistency | Nextclade, Pangolin, custom QC pipelines |
| Phylogenetic inference | Model selection, root choice, temporal signal assessment | IQ-TREE, BEAST, treetime |
| Reporting | Lineage classification, variant annotation, public health context | Lineage tools, custom reports, dashboards |

## Sampling and Sequencing Considerations

The quality of your genomic surveillance begins at the bedside or field site. Specimen type directly affects viral load and sequencing success. For respiratory viruses, nasopharyngeal swabs are routine, but for other pathogens like mpox virus, lesion swabs or scabs may be required (see [<a href="#ref-1">1</a>] for a discussion of decentralized testing approaches in Africa). Always collect metadata such as collection date, geographic location, and clinical presentation at the time of sampling.

Sequencing strategy depends on viral genome size, genetic diversity, and the surveillance question. Targeted amplicon sequencing (e.g., ARTIC protocol for SARS-CoV-2) provides high depth and is cost effective when the virus is known. Metagenomic sequencing, as used in the detection of neuroinvasive Sindbis virus from cerebrospinal fluid ([<a href="#ref-2">2</a>]), can capture unexpected or novel viruses but requires deeper sequencing and more computational filtering. For outbreak investigations, the [NCBI Sequence Read Archive](https://www.ncbi.nlm.nih.gov/sra) serves as a critical repository for depositing and accessing raw data, enabling retrospective analysis and comparison.

## Metadata: The Underappreciated Backbone

A viral genome sequence without metadata is nearly useless for surveillance. Minimum required fields include collection date, geographic location (at least administrative level), host species, and specimen type. The [NCBI Bookshelf](https://www.ncbi.nlm.nih.gov/books/) provides guidance on BioSample attributes, and many public health agencies have adopted the GISAID metadata template for influenza and SARS-CoV-2. For veterinary surveillance, as demonstrated in a genomic characterization study of Mycoplasma bovis, Mannheimia haemolytica, and bovine viral diarrhea virus 1 from a bovine respiratory disease case ([<a href="#ref-3">3</a>]), metadata on herd history, vaccination status, and clinical signs were essential for interpreting the genetic relationships.

Standardize date formats to ISO 8601 (YYYY-MM-DD). Use controlled vocabularies or ontologies for symptoms, sample types, and geographic names. Avoid free text fields that cannot be parsed programmatically. Inconsistent metadata is the most common cause of failed phylogenetic analyses and misinterpreted clusters.

## Quality Assurance and Quality Control

Quality assurance must be applied at every stage, not just after sequencing. Start with positive and negative controls during extraction and library preparation. After sequencing, assess read quality using per base quality scores, GC content, and duplication levels. The [Bioconductor](https://bioconductor.org/) project includes packages like ShortRead and Rsamtools for programmatic QC in R, while the [Galaxy Training Network](https://training.galaxyproject.org/) provides interactive workflows for these steps.

Consensus genome quality depends on depth of coverage. A minimum of 10x coverage per base is often considered the lower bound for reliable base calling, but many surveillance programs use 20x or 30x for public health decisions. Check for mixed base calls (i.e., iupac ambiguity codes) which may indicate mixed infections or sequencing errors. The emergence of an E.4 lineage in mpox virus and structural analysis of the A28L protein ([<a href="#ref-4">4</a>]) highlights how careful variant calling at the consensus level is needed to distinguish true mutations from artifacts.

## Phylogenetic Analysis and Cluster Interpretation

Phylogenetic inference places your sequences in the context of known diversity. For rapid surveillance, maximum likelihood approaches using IQ-TREE or FastTree are common. For more detailed temporal analysis, Bayesian methods (e.g., BEAST) can estimate rates of evolution and dates of common ancestry. The systematic evaluation of SNP based genotyping methods in varicella zoster virus molecular epidemiology ([<a href="#ref-5">5</a>]) demonstrates how different phylogenetic approaches can yield consistent cluster definitions when applied carefully.

A common mistake is to interpret a phylogenetic cluster as evidence of direct transmission. Phylogenetic trees show shared ancestry, not transmission routes. A cluster of identical or near identical genomes may reflect rapid spread within a community, but it could also result from a common, unsampled intermediate. An internal guide titled [Viral Phylogenetics: Interpreting Clusters Without Overstating Transmission](/blog/guides/viral-phylogenetics-interpreting-clusters-without-overstating-transmission) expands on this point. Always combine phylogenetic data with epidemiological contact information and temporal metadata to avoid overinterpretation.

## Public Health Communication and Reporting

The final step is translating genomic data into actionable information for public health decision makers. Reports should include lineage or clade assignment, key amino acid changes (e.g., in spike protein for SARS coronaviruses), and comparison with circulating strains. The rapid characterization of the 2026 Bundibugyo virus disease outbreak in the Democratic Republic of the Congo ([<a href="#ref-6">6</a>]) emphasizes that timely sharing of genomic data with global databases and local health authorities is critical for coordinated response.

Use standardized lineage nomenclature (e.g., Pango for SARS-CoV-2, Nextstrain clades for influenza) to avoid confusion. Visualizations such as time resolved phylogenies or geographic maps help communicate trends. Always state limitations: sampling bias, sequencing errors, and incomplete lineage representation. A transparent report includes a quality metrics table showing genome completeness, coverage depth, and any regions with ambiguous calls.

## Common Mistakes and How to Avoid Them

- **Ignoring sampling bias**: If all samples come from a single hospital, your phylogeny reflects local circulation, not national diversity. Stratified sampling across geography and time is essential.
- **Using default tool parameters without justification**: Default settings may not suit your viral genome (e.g., very short reads, high diversity). Document why each parameter value was chosen. The [Nextflow for Bioinformatics: Building a Reproducible Workflow](/blog/guides/nextflow-for-bioinformatics-building-a-reproducible-workflow) and [Snakemake for Research Pipelines: A Practical Starting Framework](/blog/guides/snakemake-for-research-pipelines-a-practical-starting-framework) guides provide frameworks for parameter documentation.
- **Failing to track tool versions**: Bioinformatics tools update frequently. A workflow that worked in January may produce different results in July. Use environment managers like Conda (see [Conda Environments for Bioinformatics: Managing Tools Without Version Drift](/blog/guides/conda-environments-for-bioinformatics-managing-tools-without-version-drift)) or containerization to lock tool versions.
- **Overinterpreting phylogenetic clusters**: As noted above, do not equate clustering with transmission. Add epidemiological data before drawing conclusions.
- **Neglecting provenance**: Document every file, command, and parameter. The [Reproducible Bioinformatics: Files, Environments, Parameters, and Provenance](/blog/guides/reproducible-bioinformatics-files-environments-parameters-and-provenance) guide offers practical advice.

## Limits and Uncertainty

Viral genomic surveillance has inherent limits. Sequencing errors can introduce spurious variants, especially in low coverage regions. Recombination in some viruses (e.g., norovirus, influenza) can confound phylogenetic tree inference because different genomic regions have different evolutionary histories. Sampling bias is almost always present: unsampled cases can create apparent transmission clusters that disappear when those cases are added later.

Variants may arise from laboratory contamination rather than true biological change. Include negative controls in every sequencing run and use contamination screening tools. Temporal signal in phylogenies can be weak when samples span a short time period. Before using a molecular clock, test for sufficient temporal structure (e.g., regression of root to tip distances against sampling dates). Finally, remember that genomic data alone cannot prove causality in transmission. It provides supporting evidence that must be weighed with other epidemiological information.

## Frequently Asked Questions

**What is the minimum sequencing depth for consensus calling?** A common threshold is 10x coverage per base, but many public health programs require 20x or 30x to confidently call variants. For amplicon based methods, uniform coverage across the genome is more important than average depth.

**How should I handle mixed infections or coinfections?** Mixed infections may produce ambiguous base calls or multiple variants at a single position. Use software designed for within host variant detection and report minority variants above a frequency threshold (e.g., 5%). If mixtures are suspected, consider single genome amplification or deep sequencing with unique molecular identifiers.

**Can I use a phylogenetic tree to prove that person A infected person B?** No. Phylogenetic trees show genetic relatedness, not direction of transmission. Even with very closely related sequences, there are many possible scenarios including an unsampled common source. Combine phylogenetics with detailed epidemiological interviews and contact tracing to build a stronger case.

**How often should I update the reference genome used for mapping?** Use the most recent reference genome recommended by the global surveillance community. For rapidly evolving viruses like SARS-CoV-2, reference sequences are updated periodically. Outdated references can cause mapping errors and missed variants. Check repositories such as GISAID or GenBank quarterly for new reference versions.

## Related Clinical & Scientific Guides

* [Observational vs. Experimental Studies: How to Tell Them Apart](/blog/guides/observational-vs-experimental-studies-how-to-tell-them-apart)
* [Astrocyte Single Cell Rna Seq](/blog/guides/astrocyte-single-cell-rna-seq)
* [Structural Genes](/blog/guides/structural-genes)

## References and Further Reading

<a id="ref-1"></a>[<a href="#ref-1">1</a>] [Decentralizing mpox testing in Africa](https://pubmed.ncbi.nlm.nih.gov/42434479/)

<a id="ref-2"></a>[<a href="#ref-2">2</a>] [Molecular detection of neuroinvasive Sindbis virus by NGS](https://pubmed.ncbi.nlm.nih.gov/42434987/)

<a id="ref-3"></a>[<a href="#ref-3">3</a>] [Genomic characterization of bovine respiratory disease pathogens](https://pubmed.ncbi.nlm.nih.gov/42436381/)

<a id="ref-4"></a>[<a href="#ref-4">4</a>] [Mpox lineage E.4 and A28L structural analysis](https://pubmed.ncbi.nlm.nih.gov/42434701/)

<a id="ref-5"></a>[<a href="#ref-5">5</a>] [Varicella zoster SNP based genotyping in China](https://pubmed.ncbi.nlm.nih.gov/42433640/)

<a id="ref-6"></a>[<a href="#ref-6">6</a>] [2026 Bundibugyo virus outbreak review](https://pubmed.ncbi.nlm.nih.gov/42434018/)

Related Articles

- [Viral Phylogenetics: Interpreting Clusters Without Overstating Transmission](/blog/guides/viral-phylogenetics-interpreting-clusters-without-overstating-transmission)
- [Nextflow for Bioinformatics: Building a Reproducible Workflow](/blog/guides/nextflow-for-bioinformatics-building-a-reproducible-workflow)
- [Snakemake for Research Pipelines: A Practical Starting Framework](/blog/guides/snakemake-for-research-pipelines-a-practical-starting-framework)
- [Reproducible Bioinformatics: Files, Environments, Parameters, and Provenance](/blog/guides/reproducible-bioinformatics-files-environments-parameters-and-provenance)
- [Conda Environments for Bioinformatics: Managing Tools Without Version Drift](/blog/guides/conda-environments-for-bioinformatics-managing-tools-without-version-drift)