VirSorter: mining viral signal from microbial genomic data.

Area of Science:

• Microbiology
• Virology
• Bioinformatics

Background:

• Microbial viruses are crucial in ecosystems, yet their diversity is poorly understood due to limited reference genomes.
• New genomic data types, like fragmented sequences and single-cell amplified genomes (SAGs), present challenges for viral detection.
• Existing tools for prophage detection are insufficient for fragmented and large-scale microbial datasets.

Purpose of the Study:

• To develop VirSorter, a computational tool for detecting viral signals in diverse microbial sequence data.
• To enable the identification of novel viruses from fragmented genomes, SAGs, and metagenomic data.
• To provide a scalable solution for analyzing large viral datasets and sorting viral from cellular DNA.

Main Methods:

• VirSorter utilizes probabilistic models and extensive virome data for reference-dependent and independent viral detection.
• It is designed to analyze fragmented genomic data, including assembled sequence contigs as short as 3kb.
• The tool was tested for its performance in prophage prediction and identification of viral sequences in various genomic contexts.

Main Results:

• VirSorter demonstrates comparable prophage prediction to existing tools on complete genomes.
• It excels at identifying viral sequences outside host genomes and in fragmented datasets, outperforming current methods.
• VirSorter achieves high accuracy (95% Recall, 100% Precision) on contigs >= 10kb and scales effectively for large-scale analyses.

Conclusions:

• VirSorter effectively detects viral signals in fragmented and novel microbial genomic data, addressing limitations of existing software.
• The tool facilitates the discovery of new viruses and enhances the study of uncultivated viral communities across ecosystems.
• VirSorter's availability via the iPlant Cyberinfrastructure promotes broader accessibility and application in microbial genomics research.