Viromics produces millions of viral genomes and fragments annually,
overwhelming traditional sequence comparison methods. We introduce
Vclust, an approach that determines average nucleotide identity by
Lempel-Ziv parsing and clusters viral genomes with thresholds endorsed
by authoritative viral genomics and taxonomy consortia. Vclust
demonstrates superior accuracy and efficiency compared to existing
tools, clustering millions of genomes in a few hours on a mid-range
workstation.