Petabase-scale sequence alignment catalyses viral discovery

Public sequence data represents a major opportunity for viral discovery, but its exploration has been inhibited by a lack of efficient methods for searching this corpus, which is currently at the petabase scale and growing exponentially. To address the ongoing pandemic caused by Severe Acute Respiratory Syndrome Coronavirus 2 and expand the known sequence diversity of viruses, we aligned pangenomes for coronaviruses (CoV) and other viral families to 5.6 petabases of public sequencing data from 3.8 million biologically diverse samples. To implement this strategy, we developed a cloud computing architecture, Serratus, tailored for ultra-high throughput sequence alignment at the petabase scale. From this search, we identified and assembled thousands of CoV and CoV-like genomes and genome fragments ranging from known strains to putatively novel genera. We generalise this strategy to other viral families, identifying several novel deltaviruses and huge bacteriophages. To catalyse a new era of viral discovery we made millions of viral alignments and family identifications freely available to the research community. Expanding the known diversity and zoonotic reservoirs of CoV and other emerging pathogens can accelerate vaccine and therapeutic developments for the current pandemic, and help us anticipate and mitigate future ones.