VirusTaxo

VirusTaxo classifies viral genome sequences into hierarchical taxonomic ranks using a machine learning architecture based on k-mer enrichment.

Key Features:

• Multi-class hierarchical classification: Implements a machine learning architecture tailored for multi-class hierarchical classification across viral taxonomic ranks.
• Top-down classification: Employs a top-down approach that determines higher-level ranks (e.g., order) before assigning lower-level ranks (e.g., family, genus).
• K-mer extraction and enrichment: Extracts k-mers from viral genome sequences and performs k-mer enrichment to capture discriminative sequence patterns.
• Bag-of-k-mers representation: Constructs a "bag-of-k-mers" for each class within a taxonomic rank to model class-specific genetic signatures.
• High accuracy: Achieves average accuracies for DNA viruses of 99% (order), 98% (family), and 95% (genus), and for RNA viruses of 97% (order), 96% (family), and 82% (genus).
• Novel-sequence support: Can classify novel viral sequences using full-length genomes or contigs.

Scientific Applications:

• Epidemiology: Supports tracking virus evolution and spread by providing taxonomic classification of viral genomes.
• Pathogen discovery: Aids identification and taxonomic placement of novel pathogens from genomic data.
• Vaccine development: Provides taxonomic context of viral families or genera relevant for vaccine target selection.

Methodology:

Uses a machine learning architecture for multi-class hierarchical classification based on extracted and enriched k-mers, constructs bag-of-k-mers per taxonomic class, and applies a top-down classification strategy.