Anthropic’s Life Sciences Team Reports on Six Drug Development Partnerships
Six scientists at Generate Biomedicines used Claude Code to propose new designs for one section of the genetic instructions in a CAR T-cell therapy. One design came out about 70% shorter than the standard version. In lab tests, immune cells carrying it made roughly 10 times as many receptors, the proteins cells use to spot cancer, killed tumor cells more efficiently, and stayed active longer. Generate wants the shorter section to free up room for extra instructions that would let the therapy target solid tumors. CAR T-cell therapy has historically treated blood cell cancers.
That result is one of several in a report from Anthropic’s life sciences team, which spent several months working with BigHat Biosciences, Bristol Myers Squibb, Divergence Labs, Enveda, Generate Biomedicines and The Column Group. The team also started drug discovery efforts of its own. The programs cover food allergies, rare childhood diseases and several forms of cancer, and they touch different stages of development, from molecule design to regulatory paperwork.
Antibody design at Divergence Labs
Divergence Labs is developing a food allergy treatment that targets IgE, the molecule that triggers allergic reactions. Some IgE floats freely in blood and some is attached to immune cells. Standard medications neutralize the circulating portion. Divergence wants a drug that pulls IgE off the immune cells as well, which requires a molecule with two arms. One arm must bind IgE tightly, and the other must help pry it from the cell.
Claude turned Divergence’s goals into a detailed design plan and then, largely autonomously, generated roughly 35,000 antibody designs in about a month. Early tests found hits for both arms. Three designs bound IgE and 20 dislodged it. Three of those 20 appeared to dislodge IgE more effectively than the standard medication. Divergence’s next step is to build a single drug that combines the two arms, so the combined molecule has not yet been made or tested.
Lead optimization at BigHat
BigHat Biosciences pairs AI with an automated lab. Its candidate drugs, called leads, have to bind their target tightly, remain stable through manufacturing and storage, and avoid clumping. BigHat scientists and an Anthropic researcher worked with Claude on two antibody fragments, one designed to target cancer cells and one designed to keep a drug in the body longer. Claude was connected to BigHat’s lab data system, Reccy, and proposed changes to each molecule. The lab made and tested about 1,010 variants over about 10 weeks.
Partway through, the team suspected that the standard heat-stability and clumping metrics were misleading. Claude and the researcher reanalyzed the raw data and proposed different metrics along with a change to the lab protocol. BigHat’s scientists refined and ran it, and the main predictions held. Compared with the original molecules, the best variants bound more than 10 times more tightly. The cancer-targeting lead stayed in solution at a temperature 18°C higher, and the lead for longer-lasting drugs clumped more than 100 times more slowly.
Paperwork at Enveda
At Enveda, manufacturing chemists spent up to 75% of their time on paperwork. Clinical trials require records showing each batch of a drug was made to strict quality standards, and regulators require applications built from that documentation. Enveda built a system that connects its manufacturing documents to Claude and has the model flag possible issues.
Across 1,061 pages of test documentation, Claude caught all 11 issues that human reviewers had found and flagged seven more that they had missed. Human review time fell by 89%, from over 25 hours to under three. Separately, Claude wrote first drafts of two regulatory documents. One draft took four hours, compared with about three months by hand. Enveda’s standard human review of the other draft found one minor error.
Company creation and trial planning
The Column Group is establishing a company called Pleiades, which will use Claude to run more drug programs, and more complex ones, than a small team usually could. The reasoning is that investors in biotechnology have favored previously validated ideas because discovery is slow, expensive and prone to failure. If Claude cuts the scientific and administrative overhead, riskier ideas might become easier to fund. Pleiades is still building tools and workflows. Its first Claude skill, which helps assess whether a drug target is worth pursuing, has been evaluated on 45 target-disease pairs.
Bristol Myers Squibb joined more recently. It is using Claude to assess the plan for an early-stage drug in clinical trials and to estimate the likelihood that those trials succeed. Michael Pourdehnad, the company’s head of early clinical development for hematology, oncology and cell therapy, said a single prompt produced an initial analysis that would otherwise take days, and that the work kept sensitive data secure. The two organizations plan to expand the collaboration in 2027 to see whether it can raise the probability of success in early development.
Anthropic’s own programs
Anthropic also acquired Mithril Medicine, a one-person company developing treatments for two rare forms of muscular dystrophy. Its founder, Cody Tranbarger, used Claude to organize contracts, logistics and experiments with 11 external labs, and handed other operational tasks to the model. Mithril took a possible treatment for one form of muscular dystrophy from an idea to promising results in mice in less than 18 months. Tranbarger now works at Anthropic and plans to start several new programs. If his drug candidates continue to hold up, Anthropic expects one could be ready to start human trials as soon as late 2028.
Limits and next steps
Anthropic describes the work as early and mostly preclinical, meaning the drugs have not yet been tested in people. The company says it plans to extend Claude into clinical trials as programs advance and as it partners with organizations already at the clinical stage. It reports that it has already prioritized specific model improvements based on what it learned from these programs.
Several access programs run alongside the research. Anthropic has released Claude Science, a workbench for researchers, and is offering Claude credits to life scientists. Its Life Sciences Verification Program lets trusted organizations use its most powerful models under more permissive versions of its safeguards, which can otherwise impede legitimate biological research. An open call for guest investigators invites research groups to propose projects on drug discovery questions. Organizations focused on drug discovery can contact the company through the form at the end of Anthropic’s post, which also links to technical reports from Generate, Divergence, BigHat, Enveda and the Mithril program.
