Anthropic Research has published findings on how its Claude models can assist life scientists in accelerating research, specifically in protein design and analytical chemistry. The research highlights Claude's ability to perform tasks that historically required significant time and specialized expertise.
In one experiment, Claude designed protein binders, a critical step in creating protein-based drugs. In another, Claude Opus 5 processed analytical chemistry data, matching laboratory results in minutes.
Key Points
- Claude Mythos Preview and Opus 4.8 designed protein binders against 15 targets, succeeding against 14 of them.
- Individual designs from Claude had a successful binding rate between 22% and 35%, compared to a typical 10-15% in current protein design campaigns.
- Claude Opus 5 processed Nuclear Magnetic Resonance (NMR) and Liquid Chromatography-Mass Spectrometry (LC-MS) data, returning finished results in 23 and 19 minutes.
- Claude Opus 5's analytical chemistry results matched a contract lab's analysis on hydrogen counts and purity (96.4% versus 96.33%).
- The protein design experiments involved external evaluators Adaptyv Bio and Twist Bioscience independently producing and testing Claude's designs.
- Mythos Preview achieved a 35.1% hit rate when designing against single targets separately using multiple 24-hour sessions.
Context
According to Anthropic Research, these experiments demonstrate how Claude can reduce the time and computational expertise needed for complex scientific tasks. The tasks are representative of early-stage drug development work. While life science research tasks are currently blocked in their most capable model, Anthropic states that launching an access program for scientists is a high priority.
Why It Matters
This research indicates a potential for AI models like Claude to significantly reduce the time and resource investment required for foundational scientific processes in drug discovery. Builders and researchers in life sciences may find these capabilities relevant for accelerating early-stage development and improving efficiency.
What To Do
- Note the specific Claude models (Mythos Preview, Opus 4.8, Opus 5) used for different tasks.
- Observe the reported success rates and time savings compared to traditional methods.
- Watch for announcements regarding an access program for scientists to utilize Claude's capabilities in life science research.