hiyata/herpesvirales-virus-dataset

Herpesvirales Labeled Subset

Dataset Summary

This dataset is a curated, labeled subset of hiyata/Virus-Host-Genomes, filtered to include only complete herpesvirus genome sequences from the order Herpesvirales. Sequences from Alloherpesviridae and Malacoherpesviridae were sourced directly from NCBI and appended to extend coverage across all three families in the order.

Filtering followed the same methodology described in the original dataset's accompanying publication, excluding partial sequences, mutant strains, unverified records, BAC clones, and applying a minimum genome length threshold of 50,000 bp. Four herpesvirus-specific annotation columns not present in the source dataset were added: gc_content, cpg_oe_ratio, latency_site, and cell_tropism_breadth.

Last Updated: March 30, 2026

If you use this dataset, please cite:

@article{carbajo2026sequence,
  author    = {Carbajo, Alan L and Vensko, Taylor A and Pellett, Philip E},
  title     = {Sequence Based Virus Host Prediction: A Curated Dataset and Generalizable Framework for Training Artificial Intelligence to Identify Viruses of Humans},
  journal   = {Virus Evolution},
  year      = {2026},
  pages     = {veag009},
  publisher = {Oxford University Press},
  doi       = {10.1093/ve/veag009},
  url       = {https://doi.org/10.1093/ve/veag009}
}

---

Source Dataset

This is a labeled subset of hiyata/Virus-Host-Genomes. Users interested in broader viral diversity beyond Herpesvirales should refer to the source dataset directly.

---

Motivation

Herpesviruses are defined by two biological properties that make them uniquely suited for sequence-level analysis: strict host specificity and lifelong latency. Every herpesvirus establishes latency in a characteristic cell type — sensory neurons, memory B cells, monocytes, or T cells depending on the virus — and this latency reservoir is the primary determinant of pathogenesis, reactivation risk, and zoonotic potential.

The order spans an extraordinary range of genomic properties. GC content ranges from 30.6% to 78.4% across this subset alone, and CpG dinucleotide suppression varies widely, reflecting co-evolutionary adaptation to different host methylation environments. These features are derivable from sequence but are not compiled alongside curated latency annotations in any existing resource. This subset was built to close that gap.

---

Data Fields

Field	Type	Description
sequence	string	Complete genome sequence (ACGT only)
accession	string	NCBI accession number
virus_name	string	Full virus name from NCBI taxonomy
family	string	Orthoherpesviridae / Alloherpesviridae / Malacoherpesviridae
subfamily	string	Alphaherpesvirinae / Betaherpesvirinae / Gammaherpesvirinae
genus	string	Taxonomic genus
host	string	human / non-human
standardized_host	string	Standardized host scientific name
host_category	string	Mammal / Bird / Fish / Amphibian / Mollusk
host_order	string	Host taxonomic order (e.g. Primates, Artiodactyla, Cypriniformes)
isolation_source	string	Tissue or sample source from GenBank record
isolation_date	string	Collection date
strain_name	string	Strain or isolate identifier
location	string	Geographic location of isolation
standardized_location	string	Country-level standardized location
genome_length	int32	Genome length in base pairs
gc_content	float32	Fraction G+C (0.0–1.0)
cpg_oe_ratio	float32	Observed / expected CpG dinucleotide ratio
latency_site	string	Cell type or tissue where latency is established
cell_tropism_breadth	string	broad / narrow / unknown
gemini_annotated	bool	Whether Gemini AI was used for annotation

Added Columns

The following columns were added to the source data and are not present in hiyata/Virus-Host-Genomes:

Field	Description
genome_length	Computed from the filtered sequence records
gc_content	Computed directly from each genome sequence
cpg_oe_ratio	Observed CpG dinucleotide frequency divided by expected frequency
latency_site	Curated from primary literature; subfamily-level defaults applied to uncharacterized strains
cell_tropism_breadth	Curated annotation of host cell range during lytic infection

These annotations capture the highly host-specific biology that distinguishes Herpesvirales from other viral orders.

---

Key Computed Features

GC content ranges from 30.6% to 78.4%. Betaherpesvirinae (HCMV lineage) trend toward high GC (~55–60%), while Alloherpesviridae are notably AT-rich.

CpG O/E ratio ranges from 0.229 to 1.513. Low CpG O/E ratios indicate suppression relative to random, a signature of long-term adaptation to vertebrate hosts where unmethylated CpG dinucleotides trigger innate immune sensing via TLR9.

Latency site annotations were drawn from a curated lookup table covering all well-characterized herpesviruses, with subfamily-level defaults applied to less-characterized strains.

---

Latency Biology Notes

Latency site is the primary biologically meaningful tropism annotation in this dataset, not tissue tropism. Many herpesviruses, particularly betaherpesviruses like CMV and gammaherpesviruses like EBV, infect a broad range of cell types during lytic replication and cannot be meaningfully characterized by lytic tropism alone. The latency reservoir is the clinically and evolutionarily relevant constraint.

Subfamily	Characteristic Latency Site
Alphaherpesvirinae	Sensory ganglion neuron (trigeminal, dorsal root, sacral) ¹
Betaherpesvirinae	Monocyte / CD34+ hematopoietic progenitor ²
Gammaherpesvirinae	Resting memory B lymphocyte (or T lymphocyte in some) ³
Alloherpesviridae	Largely unknown; possibly leukocytes in fish herpesviruses ⁴
Malacoherpesviridae	Unknown; hemocytes suspected in OsHV-1 ⁵

¹ Gilden et al. (2001). Presence of VZV and HSV-1 DNA in Human Nodose and Celiac Ganglia. Virus Genes, 23, 145–147. https://doi.org/10.1023/A:1011883919058

² Roizman B., Knipe D.M., Whitley R.J. Herpes Simplex Viruses. In: Knipe D.M., Howley P.M., editors. Fields Virology. 6th ed. Volume 2. Lippincott Williams & Wilkins; Philadelphia, PA, USA: 2013.

³ Longnecker R., Kieff E., Cohen J.I. Epstein–Barr Virus. In: Knipe D.M., Howley P.M., editors. Fields Virology. 6th ed. Volume 2. Lippincott Williams & Wilkins; Philadelphia, PA, USA: 2013.

⁴ Reed AN, Izume S, Dolan BP, et al. Identification of B cells as a major site for cyprinid herpesvirus 3 latency. J Virol. 2014;88(16):9297–9309. https://doi.org/10.1128/JVI.00990-14

⁵ Divilov K, Wang X, Swisher AE, et al. Ostreid herpesvirus 1 latent infection and reactivation in adult Pacific oysters, Crassostrea gigas. Virus Res. 2024;339:199245. https://doi.org/10.1016/j.virusres.2023.199245

---

Filtering Methodology

Sequences were filtered from hiyata/Virus-Host-Genomes using the same criteria described in the accompanying publication: partial sequences, mutant strains, unverified records, and BAC clones were excluded, and a minimum genome length threshold of 50,000 bp was applied to remove fragments. Only records classified within the order Herpesvirales were retained.

---

Limitations

• Latency site annotations for Alloherpesviridae and Malacoherpesviridae are based on limited literature and should be treated as provisional.
• Sampling reflects NCBI submission patterns. Well-studied human herpesviruses and economically important animal herpesviruses (koi herpesvirus, Marek's disease virus) are overrepresented relative to wildlife herpesviruses.
• Records receiving subfamily-level defaults for latency annotations should be interpreted with appropriate uncertainty, particularly for novel or poorly characterized strains.