Towards Standards in Viromics: in silico Evaluation of Viral Identification, Taxonomy, and Auxiliary Metabolic Genes (AMGs) Curation

Abstract:

Viruses influence microbial diversity and global nutrient cycles, and metagenomic approaches have been critical for revealing viral roles across diverse ecosystems. However, the emergent field of viral ecogenomics has many processes that would benefit from comparative benchmarking to better standardize and enable comparison across datasets. Here we evaluate (i) viral identification, (ii) viral classification, and (iii) viral auxiliary genes (AMGs) curation using in silico-generated datasets that mimicked features of viromes and bulk metagenomes to provide guidelines and highlight potential pitfalls of viral metagenomic analyses.

Diverse viral identification tools are now available, e.g. VirSorter, MARVEL, MetaPhinder, deepVirFinder and VIBRANT. The results of our benchmarking show that tools based on gene content including VirSorter, MARVEL and VIBRANT, consistently outperformed other tools except for small (<3kb) viral contigs. Though tools such as, MetaPhinder, and DeepVirFinder outperformed these for small contigs, they did so at the cost of a higher false positive rate, particularly when eukaryotic or mobile element sequences were included in test datasets. For viral classification, variously sized genome fragments were assessed using gene-sharing network analytics (vConTACT2) for concordance against known taxonomy. Taxonomic classification of viral contigs was found to be acceptable (37.5% correct assignation) for contigs >3kb and, quickly improved as fragment length increases (~50% correct assignation for contigs >10kb). Finally, we outline suggestions and best practices for researchers to manually inspect and validate candidate auxiliary metabolic genes (AMGs) in metagenome-assembled viral genomes.

Together these benchmarking experiments provide guidance for researchers seeking to best detect and characterize the myriad viruses ‘hidden’ in diverse sequence datasets.