Claude Cuts CAR-T Therapy Component by 70% in Drug Discovery Push

Claude's move into real drug discovery

Anthropic has laid out how Claude is being used across the messy, expensive middle of drug development — not just writing summaries, but designing molecules and shaping clinical strategy.

The company detailed collaborations with several biotech firms and one pharma giant in a blog post this week. The projects span cancer, food allergies and rare muscular disorders, though most treatments remain experimental.

The headline result comes from Generate Biomedicines, where six scientists used Claude Code to redesign a component of CAR T-cell therapy. The approach, traditionally used against blood cancers, was modified to leave more room for genetic instructions that could help it tackle solid tumors.

The redesigned component came out roughly 70% shorter. In lab experiments, the modified immune cells produced about ten times more cancer-detecting receptors, destroyed cancer cells more effectively, and stayed active longer, according to the reported findings.

Antibodies, manufacturing and trial predictions

Beyond the CAR-T work, Claude is being used to generate antibody candidates, optimize manufacturing documentation and even estimate whether clinical trials will succeed.

Divergence Labs turned Claude loose on antibody design for a food-allergy treatment targeting IgE. The system produced about 35,000 molecular designs in a month. Initial tests found three designs that bind IgE and 20 that detach it from immune cells — and three of those 20 looked more effective than the standard medication at removing IgE.

BigHat Biosciences paired Claude with automated lab equipment to improve two antibody fragments. Over ten weeks, researchers made and tested about 1,010 variants. The strongest cancer-targeting version bound its target more than 10 times as tightly and stayed stable at temperatures 18°C higher than the original. The other fragment aggregated more than 100 times more slowly — a big deal for longer-lasting medicines.

Enveda used Claude on manufacturing documentation, a task that had eaten up to 75% of some chemists' working time. In a 1,061-page assessment, the system caught all 11 problems human reviewers had found — plus seven more. Manual checking time dropped 89%, from over 25 hours to under three.

Bristol Myers Squibb bets on Claude for clinical trials

The most commercially significant partnership is with Bristol Myers Squibb, which is using Claude to assess clinical study plans and estimate their chances of success.

Michael Pourdehnad, BMS senior vice president for early clinical development, said in a statement: "From a single prompt, Claude generated a thorough initial analysis." He noted this helped evaluate development options that would otherwise have taken several days.

The companies plan to extend the work into early clinical development decisions. For Anthropic, the collaboration is a two-way street: working directly with drug developers exposes weaknesses in its models and points to improvements across scientific research.

What it means for you

For readers in the US, UK and EU, these collaborations are early-stage research projects, not approved treatments. None of the drug candidates described are available to patients, and the results are lab findings that still need clinical validation.

Anthropic has not announced consumer-facing changes to Claude based on this work. The partnerships are with biotech and pharma companies, and any new tools for life-science organizations would likely be enterprise offerings. Pricing and availability for such tools have not been disclosed.

Anthropic's pitch is that Claude can compress the most expensive parts of drug development — from molecular design to trial planning. The early numbers are impressive, but they come from lab experiments and a single pharma partner. The next step is proof in real clinical programs.