vRhyme

vRhyme bins viral genomes from metagenomic datasets to generate high-quality viral metagenome-assembled genomes (vMAGs) for analysis of viral diversity and function.

Key Features:

• Coverage effect size comparisons: Performs single- and multi-sample coverage effect size comparisons between scaffolds to distinguish patterns indicative of distinct viral genomes.
• Supervised machine learning and weighted networks: Uses supervised machine learning to identify nucleotide feature similarities and to construct weighted networks that are iteratively refined to form genome bins.
• Protein redundancy scoring: Incorporates a protein redundancy scoring mechanism based on the expectation that viral genomes typically do not encode redundant genes to improve bin accuracy.

Scientific Applications:

• Benchmarking with simulated viromes: Demonstrated reconstruction of more complete and less contaminated vMAGs compared to existing binning tools on simulated virome datasets.
• Human skin virome analysis: Applied to 10,601 viral scaffolds from human skin metagenomes to bin complex viral sequence collections.
• vMAG discovery examples: Recovered a Herelleviridae vMAG composed of 22 scaffolds and a vMAG encoding a nitrate reductase metabolic gene, representing near-complete genomes after binning.

Methodology:

Constructs weighted networks from nucleotide features identified via supervised machine learning that are iteratively refined to produce genome bins, and uses single- and multi-sample coverage effect size comparisons plus protein redundancy scoring to guide bin refinement.