'Google for DNA' brings order to biology's big data (www.nature.com)

Researchers have released MetaGraph, a graph‑based search engine for raw biological sequencing data described in Nature that aims to be a “Google for DNA.” MetaGraph integrates seven public repositories into a compressed, queryable index that covers 18.8 million unique DNA/RNA sequence sets and 210 billion amino‑acid sequences across viruses, bacteria, fungi, plants and animals. Using mathematical sequence graphs that link overlapping fragments, the system lets users submit textlike prompts to retrieve genetic patterns hidden in petabases of noisy, fragmented reads—patterns that aren’t explicitly annotated—enabling searches that were previously infeasible at this scale. For the AI/ML community this is significant because it transforms inaccessible bulk sequencing archives into an on‑the‑fly, searchable knowledge base suitable for large‑scale analyses and model training. The authors demonstrate practical performance by scanning 241,384 human gut microbiomes for antibiotic‑resistance signals in about an hour on a high‑end machine, and earlier work used the tool to track resistance genes across subway systems. MetaGraph sets a new standard for compressing, indexing and retrieving raw biological data, lowering the barrier for downstream tasks such as surveillance, meta‑analyses and training sequence models—while also raising future challenges around compute, integration with ML pipelines, and responsible access to human genomic content.