HVUE: Human Virome Understanding Evaluation

Dataset Description

HVUE (Human Virome Understanding Evaluation) is a comprehensive benchmark for evaluating foundation models on viral genomics tasks. The benchmark comprises 7 curated datasets across 3 epidemiologically critical prediction tasks:

• Pathogenicity Classification (3 datasets)
• Host Tropism Prediction (1 dataset)
• Transmissibility Assessment (3 datasets)

Paper: HViLM: A Foundation Model for Viral Genomics Enables Multi-Task Prediction of Pathogenicity, Transmissibility, and Host Tropism

Authors: Pratik Dutta, Jack Vaska, Pallavi Surana, Rekha Sathian, Max Chao, Zhihan Zhou, Han Liu, and Ramana V. Davuluri

GitHub: https://github.com/duttaprat/HViLM

Dataset Structure

Pathogenicity Classification

CINI Dataset

• 159 sequences across 4 viral families
• Manual literature-based curation
• Binary classification: pathogenic vs non-pathogenic

BVBRC-CoV Dataset

• 18,066 coronavirus sequences
• Distinguishes human-pathogenic (SARS-CoV-2, MERS-CoV, etc.) from animal-restricted strains

BVBRC-Calici Dataset

• 31,089 calicivirus sequences
• Clinical evidence and isolation source-based labels

Host Tropism Prediction

VHDB Dataset

• 9,428 sequences spanning 30 viral families
• Binary classification: human-tropic (13.1%) vs non-human-tropic (86.9%)
• Experimentally validated host range annotations

Transmissibility Prediction

Coronaviridae Dataset

• ~3,000 coronavirus sequences
• R₀-based classification: R₀<1 vs R₀≥1

Orthomyxoviridae Dataset

• ~2,500 influenza sequences
• R₀-based classification

Caliciviridae Dataset

• ~1,800 calicivirus sequences
• R₀-based classification

Data Format

Each dataset contains three splits:

• train.csv
• dev.csv
• test.csv

CSV columns:

• sequence: Viral genomic sequence (250-1000 bp)
• label: Binary label (0 or 1)

Usage

from datasets import load_dataset

# Load entire benchmark
hvue = load_dataset("duttaprat/HVUE")

# Load specific task
patho_cini = load_dataset("duttaprat/HVUE", data_files="pathogenicity/CINI/*.csv")

# Load specific split
train_data = load_dataset("duttaprat/HVUE", data_files="pathogenicity/CINI/train.csv")

Citation

@article{dutta2025hvilm,
  title={HViLM: A Foundation Model for Viral Genomics Enables Multi-Task Prediction of Pathogenicity, Transmissibility, and Host Tropism},
  author={Dutta, Pratik and Vaska, Jack and Surana, Pallavi and Sathian, Rekha and Chao, Max and Zhou, Zhihan and Liu, Han and Davuluri, Ramana V.},
  journal={Submitted to RECOMB},
  year={2025}
}

License

CC-BY-4.0

Contact

• Pratik Dutta: Pratik.Dutta@stonybrook.edu
• GitHub Issues: https://github.com/duttaprat/HViLM/issues