Anthropic’s large language model Claude generated roughly 35,000 antibody designs within about one month, and initial laboratory tests identified several candidates that could be relevant to treating food allergies by targeting Immunoglobulin E (IgE).
Who is involved and the reported outcomes
- The work was conducted with Divergence Labs, a U.S. research organization that is one of six drug-development teams collaborating with Anthropic to explore how AI can shorten drug development timelines. Anthropic reported that the collaboration improved antibody stability and reduced time spent reviewing pharmaceutical documents by 89%.
- Claude produced approximately 35,000 potential molecular designs focused on a strategy to target IgE, the antibody class implicated in allergic reactions.
Therapeutic approach and molecular design
Divergence Labs is pursuing a treatment approach that differs from existing therapies, which mainly neutralize IgE circulating in the bloodstream. Their goal is to remove IgE from immune cells. To achieve this they require a bifunctional molecule: one part must bind tightly to IgE, and the other must facilitate detachment of IgE from immune cells.
Laboratory findings
In primary laboratory assays the researchers identified three designs that can bind IgE and twenty designs capable of detaching IgE from immune cells. According to Anthropic, three of the detaching designs outperformed a conventional drug in the laboratory tests. Researchers are now working to combine both functions in a single therapeutic molecule.
Related work on CAR‑T cell therapy
Anthropic is also collaborating with Generate Biomedicines on redesigning aspects of CAR‑T cell therapy. Using Claude Code, researchers redesigned genetic instructions that regulate modified immune cells. One proposed redesign reduces the process length by about 70% compared with current methods.
Lab evaluations indicate the redesigned cells produced roughly ten times as many tumor-recognizing receptors, killed cancer cells more effectively, and remained active for a longer period. These results suggest the possibility of expanding CAR‑T use beyond primarily blood cancers to other tumor types in the future.
Caveats and next steps
Researchers stress these are preliminary laboratory results and do not yet demonstrate safety or efficacy in patients. Additional in vivo studies, broader preclinical work, and clinical trials will be required before any therapeutic claims can be substantiated.
Summary
Claude’s rapid generation of tens of thousands of antibody designs produced multiple laboratory-validated IgE-targeting candidates, and AI-assisted redesigns of CAR‑T constructs show improved functional metrics in vitro. While these developments could accelerate drug discovery, clinical validation remains necessary.
