Anthropic Early progress on accelerating drug development with Claude
Accelerating Drug Development with AI Agents
Anthropic’s life sciences team is actively integrating Claude into the drug development lifecycle to address bottlenecks ranging from molecular design to regulatory compliance. By partnering with six organizations—BigHat Biosciences, Bristol Myers Squibb, Divergence Labs, Enveda, Generate Biomedicines, and The Column Group—Anthropic is demonstrating that AI can function as an autonomous agent to reduce the time and cost required to bring new medicines to patients. These efforts are currently focused on preclinical research, with plans to scale operations into clinical trial planning.
Key Case Studies and Technical Applications
Modality Engineering: Generate Biomedicines
Generate Biomedicines utilized Claude Code to redesign the signaling domain of CAR T-cell therapy. The goal was to create a more compact receptor to enable multi-functional cancer-fighting capabilities. The resulting design was 70% shorter than the standard version and demonstrated a 10-fold increase in receptor expression on the cell surface, leading to more efficient tumor cell destruction in laboratory tests.
Hit Generation: Divergence Labs
To address food allergies, Divergence Labs tasked Claude with designing a bispecific antibody capable of detaching IgE from immune cells. Claude autonomously generated approximately 35,000 antibody designs over one month. Laboratory validation identified 20 "hits" capable of dislodging IgE, with three designs outperforming the standard medication, Xolair, in early tests.
Lead Optimization: BigHat Biosciences
BigHat Biosciences combined Claude with its automated lab system, Reccy, to optimize two antibody fragments. Over 10 weeks, the team tested 1,010 variants. Beyond design, Claude and the researchers identified that standard heat-stability metrics were misleading. By reanalyzing raw data, they developed new metrics and protocols, resulting in molecules that bound 10x more tightly and showed significantly improved stability and reduced clumping.
Regulatory Documentation: Enveda
Enveda implemented a system called the CMC Document Processor (CDP) to automate the review of manufacturing records. In a test involving 1,061 pages of documentation, Claude identified all 11 issues previously found by human reviewers, plus seven additional issues. This process reduced human review time by 89%, from 25 hours to less than three. Additionally, Claude drafted portions of an FDA Investigational New Drug (IND) application in four hours, a task typically taking three months by hand.
Company Creation: The Column Group (TCG)
TCG is launching a new company, Pleiades, designed to be "AI-enabled from Day One." By using Claude to evaluate target-disease pairs, TCG has reduced the time required for initial target assessment from weeks to days. This efficiency allows for a higher volume of research programs, enabling investment in riskier, novel biological targets that were previously considered too costly to pursue.
Implications for Future Research
Anthropic’s strategy emphasizes that drug development bottlenecks are distributed throughout the entire process, not just in initial molecule design. By providing Claude credits to life scientists and launching the Life Sciences Verification Program, the company aims to lower the barrier for researchers to use powerful models in biological contexts. The ongoing work with Bristol Myers Squibb to optimize clinical trial planning marks a shift toward applying these AI capabilities to later-stage development. Anthropic’s acquisition of Mithril Medicine and its "Pipeline-in-a-Person" model further suggests a future where a single researcher, supported by AI agents, can manage multiple drug discovery programs simultaneously.