Anthropic reports the first 3D structure of a prokaryotic vault particle, found with Claude

Anthropic published a post and a bioRxiv preprint from its in-house biology lab describing what it calls the first 3D structure of a vault particle from prokaryotic cells. Vaults are large, hollow protein shells found across most of the eukaryotic tree; they were discovered in 1986, and their function is still unknown. The preprint is not peer-reviewed, and its competing-interest statement says all authors are Anthropic employees who may hold equity.

Anthropic describes the work in three steps. Using Claude Science, the team searched a database of about 15 billion predicted proteins for relatives of the major vault protein, and reported finding them in 29 bacterial and four archaeal phyla, but in under 1% of species with sequenced genomes. They tested 44 candidates in a cell-free system and moved the best into E. coli, where 15 of 24 produced soluble protein; seven of those 15 assembled into vault-like particles under the electron microscope, and cryo-electron microscopy confirmed the shape for four. For one phage protein, called L05, the structure shows a shell of 96 copies enclosing roughly 25% more volume than a human vault. Adding wall repeats to the protein produced larger vaults.

Anthropic’s account of Claude’s role is specific. Claude designed DNA constructs, recommended a size-exclusion method that cut purification from about three days to about 90 minutes, and, directed by the team, ran the cryo-EM processing. A first structural model was ready in 34 hours, which Anthropic estimates would normally take weeks. The post also records disagreement inside the process: Claude first argued against the chromatography approach, and the team went ahead regardless. These timings and the “first” claim are Anthropic’s own; the post notes that related proteins had already been found in prokaryote and phage genomes, and that it is not the first group to isolate vaults on sizing columns.

Anthropic did not determine what vaults do. It reported an unidentified small molecule in a pocket of each protein copy, whose shape Claude found resembled coenzyme A, as a possible but unconfirmed hint. Leonard Rome, who co-discovered vaults and advised the team, was quoted as saying that establishing their function remains “enormously challenging”. The structures and further data were promised “soon”.

The paper follows Anthropic’s September claim of a new enzyme system, and fits the company’s stated aim of building Claude into every step of laboratory biology.