Anthropic says Claude can now design proteins and analyse lab data

Anthropic says Claude can now do more than chat and code; it can design protein binders and analyse chemistry lab data. The company says the AI completed tasks that can take scientists hours or even weeks, hinting at a much bigger role for AI in research. Claude can design proteins and analyse lab data, Anthropic says.Anthropic just announced that Claude is doing something new, it is taking part in work that historically required scientists and weeks or months of time. In a new research update, Anthropic shared two experiments showing how Claude can help with protein design and analytical chemistry. The company says the results point to a future where AI could become a more active research assistant, handling some of the tedious computational and data-analysis work involved in scientific research.Designing proteins like a lab researcherThe first experiment involved designing protein binders from scratch. In simple terms, a protein binder is a small protein designed to attach itself to another protein. Such binding is important in drug development because many medicines work by attaching to a specific target in the body.Anthropic asked Claude to design 30 potential binders for each of 15 protein targets. The designs were then produced and tested in real laboratories by Adaptyv Bio and Twist Bioscience.The result? Claude successfully designed binders for 14 of the 15 targets, producing 354 confirmed binders from 1,320 designs. Anthropic says the models achieved overall hit rates of 22.6 per cent and 26.7 per cent in one setup, compared with the roughly 10–15 per cent typically seen in protein-design campaigns. When Mythos Preview worked on individual targets separately, its hit rate rose to 35.1 per cent.Some of the designs also showed strong binding, with Anthropic saying several matched or exceeded previously published results. That does not mean Claude has suddenly designed a new drug. Protein design is only one of the early steps in drug development, and promising designs still need extensive laboratory testing, optimisation and clinical development.Reading raw lab data in minutesAnthropic gave Claude raw data from two common chemistry techniques, NMR and LC-MS, which scientists use to identify compounds and check how pure they are. Instead of relying on specialist software and a human operator, Claude analysed the files itself.Claude processed the NMR data in 23 minutes and the LC-MS data in 19 minutes. Its analysis reported a purity of 96.4 per cent, almost identical to the laboratory's own figure of 96.33 per cent.The AI also worked with a raw LC-MS file stored in an undocumented format used by the instrument manufacturer. Anthropic says Claude figured out how the data was encoded, checked that it had interpreted the file correctly, and produced the analysis a chemist would normally expect.There is a bigger point behind these experiments. Claude is not just being used as a chatbot that explains science. It is being used to operate tools, run specialised models, analyse experimental data and carry out multi-step research tasks.Anthropic also acknowledges a catch: these capabilities are dual-use. The same AI that could speed up drug research could potentially be misused for dangerous biological work. For that reason, the company says advanced life-science research capabilities remain restricted while it works on ways to give scientists access safely.- EndsPublished On: Aug 19, 2026 14:14 IST