ViroNEXT: A multilayer hybrid system for high-precision detection of known and divergent viruses in metagenomics

Metagenomic sequencing enables unbiased virus detection but remains challenged by false-positive classifications and limited detection of highly divergent viruses. We present ViroNEXT, a fully automated end-to-end pipeline combining reference-based virus identification with complementary machine-learning-assisted detection of divergent viral sequences. Multilayer nucleotide-, protein-, assembly- and coverage-based analyses improve classification reliability. Across 40 synthetic, spike-in and clinical datasets, ViroNEXT substantially reduced false-positive assignments while achieving recall comparable to or exceeding established workflows and detecting highly divergent viruses missed by homology-based approaches. ViroNEXT provides interpretable automated reporting and is freely available as open-source software and through a public web interface, facilitating broad adoption of viral metagenomic analysis.