Anthropic reported on November 4, 2026, that its biology team used Claude to reconstruct the first 3D structure of a vault particle derived from a virus that infects bacteria. Vaults are large, hollow protein shells whose biological function remains unknown.
Using Claude Science, the researchers searched a database of approximately 15 billion predicted proteins. They found major vault protein homologs in 29 bacterial phyla and four archaeal phyla, although these proteins appeared in less than 1% of species with sequenced genomes.
The researchers selected 44 proteins for cell-free expression testing before moving leading candidates into E. coli. Fifteen of 24 candidates produced soluble protein, and seven of those 15 formed vault-like particles under electron microscopy. Cryo-electron microscopy confirmed the vault shape in four of them, two derived from phages and two from bacteria.
The phage-derived L05 particle was selected for detailed structural analysis because it was among the best at forming complete, closed particles. It contains 96 copies of one protein, arranged as 48 copies in each half, with a mass of roughly 9.5 megadaltons, compared with 78 protein copies in mammalian vaults. Claude helped the team map the shell at a nominal resolution of 3.2 angstroms.
Claude also changed the laboratory workflow. It proposed switching from a three-day ultracentrifugation protocol to a 90-minute direct size-exclusion chromatography run, and it set up and ran the open-source RELION workflow while researchers guided the analytical decisions. The team produced its first complete 3D model in 34 hours, and the entire project was completed in less than four months.
Anthropic also documented Claude's mistakes. One laboratory protocol it generated included an unsuitable additive in the bacterial growth medium, which Anthropic says set the project back by two weeks. Claude also skipped a standard step while processing microscopy images and produced an incorrect structural map that combined the particle's image with its mirror image, before manual validation caught it. Anthropic has submitted the findings as a preprint to bioRxiv, and the research has not yet been peer reviewed.
