Anthropic’s research team evaluated Claude’s performance on two representative life‑science tasks. In a protein‑design campaign, Claude Opus 4.8 and Mythos Preview generated binders for 15 protein targets; external labs Adaptyv Bio and Twist Bioscience confirmed binders for 14 of the 15 targets. Hit rates ranged from 22.6% to 35.1%, exceeding the typical 10–15% seen in current protein‑design campaigns, and some binders matched or surpassed the best published affinities. In a separate experiment, Claude Opus 5 processed raw NMR and LC‑MS data files provided by a contract lab. With only a two‑sentence prompt, the model returned complete analytical results in 19 to 23 minutes, achieving a purity of 96.4% versus the lab’s 96.33% and matching hydrogen‑count measurements. These studies illustrate that Claude can reduce the time and computational expertise required for early drug‑development tasks, potentially accelerating the overall discovery pipeline. The work was conducted with minimal human oversight, relying on Claude’s autonomous execution, and the results were validated by independent external evaluators. Anthropic plans to expand access to its models for scientists, with Opus 5 currently being the most capable generally available model.
How Claude is accelerating protein design and analytical chemistry
Anthropic’s Claude models were tested on two life‑science tasks: designing protein binders for 15 targets and analyzing NMR/LC‑MS data. In the protein design experiment, Claude succeeded on 14 of 15 targets, achieving hit rates up to 35% and producing binders with affinities equal to or better than published results. In the analytical chemistry test, Claude Opus 5 processed raw lab files in 19–23 minutes, matching the lab’s purity and hydrogen‑count results. These outcomes suggest Claude can accelerate early drug‑development stages.
Why it matters
Demonstrating that general‑purpose AI can match or surpass specialist tools in protein design and analytical chemistry could lower barriers to entry and speed drug discovery pipelines.