Benchmarking Read-Based Virome Profilers for Human Virus Detection and Community Discovery

Background: Shotgun metagenomics enables human virome profiling, yet readbased platforms differ in databases and mapping strategies, potentially altering biological conclusions, especially when goals diverge between human virus detection and broader virome discovery.Methods: We benchmarked Kraken2, FastViromeExplorer, and ViromeScan on whole-genome shotgun datasets from gut, nasal, oral, and vaginal sites.Reads were trimmed, human-depleted, and profiled; counts were normalized to CPM.We compared alpha diversity, beta diversity, and taxonomic composition.Results: Platform choice strongly affected viral composition.FastViromeExplorer produced the highest CPM and alpha diversity.Kraken2 was intermediate, and ViromeScan was lowest.Beta-diversity showed clear site clustering for all platforms, strongest with FastViromeExplorer, moderate with Kraken2, and weakest with ViromeScan.At the phylum level, Uroviricota dominated Kraken2 and FastViromeExplorer, whereas ViromeScan emphasized Nucleocytoviricota and Peploviricota.Genus and species calls were highly platform-dependent.FastViromeExplorer detected the most taxa overall, while ViromeScan preferentially reported eukaryotic and human viruses.Conclusions: Read-based profilers yield divergent virome portraits driven by database scope and mapping stringency.For clinical human-virus detection, curated human and eukaryotic references with coverage-based confirmation are essential.Our results provide practical guidance for aligning pipelines to study goals and underscore the need to report parameters, database versions, and cross-platform validations when interpreting the human virome.