The Pulse
Anthropic Says Claude Helped Map a Phage Vault Structure
Anthropic reports a 3D structure of a vault particle built from a bacteriophage protein and produced in E. coli. The company says Claude helped the team search protein sequences, isolate particles and process cryo-EM data, while the vault’s
AI.info Team · Written with AI · fact-checked ·
Anthropic’s biology research lab reports the first 3D structure of a vault particle found in prokaryotic cells. The company published the finding on Nov. 4, describing how Claude assisted with sequence searches, particle isolation and structural analysis. The team says it has not yet determined the particles’ biological function.
The particle is built from a bacteriophage protein, which the researchers call L05, and was made inside E. coli for study. Bacteriophages are viruses that infect bacteria, and their proteins are made inside host cells. The researchers say a vault in a prokaryotic cell gives researchers a simpler system to study than mammalian cells do.
The L05 vault’s 96-protein shell
The team used cryo-electron microscopy to map the structure, which contains 96 copies of L05, or 48 copies in each half. Together, those proteins have a mass of about 9.5 megadaltons. The mammalian vault has 39 copies per half, according to Anthropic.
The L05 vault measures roughly 54 nanometers long and 45 wide, compared with 67 by 40 nanometers for mammalian vaults. Its shorter, wider form encloses about 25% more volume, the company says. Seven repeating segments form the L05 vault’s wall, while mammalian vaults have nine, and the L05 end caps are flatter.
The cap closes in three steps: among each set of three neighboring protein copies, one ends at the first step, another at the second, and the third reaches the tip. The copies are identical along the wall but take different shapes in the cap. Anthropic describes the particle as among the largest protein shells identified in prokaryotes.
Claude’s role in finding and isolating particles
Claude Science searched a database of about 15 billion predicted proteins for relatives of the vault protein. Anthropic found related proteins in 29 major groups of bacteria and four groups of archaea, but in less than 1% of species with a sequenced genome. The team selected 44 proteins to test in a cell-free system.
Claude turned the selected protein sequences into DNA designs tailored to the host cell, while the researchers chose which proteins to test. The team moved its best performers into living E. coli, where 15 of 24 produced soluble protein. Under an electron microscope, seven of the 15 proteins had assembled into vault-like particles.
Cryo-EM confirmed vault shapes for four of those seven particles: two from phages and two from bacteria. The researchers chose the phage protein L05 for detailed analysis because it was among the best at forming complete, closed particles. Claude also helped the team develop a faster way to isolate the vaults.
The conventional purification method takes approximately three days and uses a tube of increasingly dense sugar. Anthropic instead used size-exclusion chromatography, with Claude recommending a column and ways to track the protein and check for particles. The optimized method takes about 90 minutes from cells to microscopy grid.
For the structural analysis, Claude set up the open-source RELION software and ran processing steps under the team’s direction. A first full model was ready in 34 hours, and the final maps were ready about a week after processing began. The cryo-EM analysis used hundreds of GPUs in parallel to save time.
Engineered shells and an unidentified molecule
The researchers made 15 altered versions of the L05 protein, most with changes where neighboring copies touch. Some single changes near the cap stopped the shell from closing, producing open, rosette-like particles instead of sealed vaults. Anthropic says the results show that this region is needed to seal the shell.
The team also copied two wall segments, increasing the protein from seven repeating segments to nine. Those engineered proteins assembled into larger vaults, with the enclosed volume increasing by approximately 20%. Anthropic says the experiments demonstrate how vaults could have changed size over their evolutionary history by duplicating pieces of their own wall without disrupting how they assemble.
A pocket in each of the 96 protein copies holds a small molecule visible in the structural map. Claude found that the molecule’s shape resembles coenzyme A, but the researchers have not identified it. The company says that if the molecule is coenzyme A, or something similar, it would be an early hint about a possible function.
Function and experimental limits
Anthropic’s team has made the vaults only in E. coli so far and has not tested whether they form in the bacteria and phages where the proteins occur naturally. The researchers say more work is needed to determine what the particles do. Vaults have been studied in eukaryotes, but their function remains unresolved.
The project also exposed mistakes in Claude’s experimental guidance: one protocol called for an additive that interfered with getting DNA into cells, setting the work back by two weeks. During early cryo-EM processing, Claude skipped a standard step and produced a flawed map blended with its mirror image. Anthropic says the team shared its lab notes with the researchers who train Claude, and that the model improved during the project.
The company says it is releasing the findings early and has submitted a preprint to bioRxiv. Anthropic plans to release the structures and additional data.