ViralMultiNet SARS-CoV-2 ORF-level Functional Classification Dataset

This dataset supports the publication "ViralMultiNet: A Structure-Aware Multimodal Framework for Viral Genome Classification" in GigaScience (2025). The dataset contains 66,011 processed SARS-CoV-2 ORF sequences (300-900 bp) with functional classifications derived from wastewater metagenomic surveillance data. Dataset Contents:- Balanced training (47,091), validation (9,417), and test (9,503) splits- Four functional classes: Enzymatic, Structural, Transport, Other- Pre-computed BERT annotation embeddings (768-dimensional)- K-mer to protein structure residue mappings- PDB structure files (6VXX, 7BV2)- High-attention k-mer regions (4-7-mer scales) Source code: https://github.com/DAVE-HUB-11/ViralMultiNet-gigascienceRaw data: NCBI BioProject PRJNA1251232License: MITContact: xwx@guet.edu.cn