Anthropic uses Claude AI to cut pharmaceutical document review times by 89%

Streamlining the paperwork of medicine

Drug development is often slowed down by massive amounts of administrative documentation. For some chemists, manufacturing documentation can consume as much as 75% of their total working time, creating a significant bottleneck in the production of new medicines.

Anthropic is addressing this by integrating Claude AI into the workflows of life-science organizations. The company recently collaborated with Enveda to assess a manufacturing documentation set that spanned 1,061 pages. The goal was to see if the AI could handle the tedious task of identifying errors and compliance issues.

The results showed that Claude identified all 11 problems previously found by human reviewers, plus seven additional issues the humans had missed. As reported by TechRadar, this approach slashed the manual checking time by 89%, bringing a process that usually takes more than 25 hours down to under three hours.

Accelerating molecular design

Beyond paperwork, Anthropic is using Claude to help scientists design the actual components of new drugs. At Generate Biomedicines, researchers utilized the AI to alter a part of a CAR T-cell therapy, which is a treatment typically used against blood cancers.

The scientists wanted to create more space for genetic instructions so the treatment could eventually target solid tumors. By using the AI to redesign the component, they produced a version that was approximately 70% shorter than the original design. This modification allowed the modified immune cells to produce roughly ten times more cancer-detecting receptors.

Laboratory experiments further showed that these AI-assisted cells destroyed cancer cells more effectively and stayed active for longer periods. This demonstrates how the model can help scientists iterate on complex biological structures to improve the potency of existing therapies.

Testing new antibody candidates

Other biotech firms are using Claude to explore a wider variety of chemical possibilities for specific diseases. Divergence Labs utilized the system to develop antibody candidates for a potential treatment for food allergies involving the immune molecule IgE.

Within a single month, the AI generated approximately 35,000 molecular designs. These designs were then tested in initial experiments to see which ones could bind to IgE or detach it from immune cells. The researchers identified three designs capable of binding to IgE and another 20 capable of detaching it from immune cells.

Similarly, BigHat Biosciences paired Claude with automated lab hardware to enhance two antibody fragments. Over ten weeks, they tested about 1,010 variations to check for binding strength and stability. The strongest cancer-targeting version bound its target more than 10 times as tightly as the original and remained stable at temperatures 18°C higher.

Predicting clinical success

Anthropic is also moving into the clinical research phase of drug development through a partnership with Bristol Myers Squibb. The companies are using Claude to assess clinical study plans and estimate the likelihood of their success before moving forward with expensive trials.

Michael Pourdehnad, a senior vice president at the pharmaceutical firm for early clinical development, noted that Claude could generate a robust initial analysis from a single prompt. This allows the team to evaluate different development options that would normally require several days of manual work by human experts.

The companies plan to expand this collaboration to see how Claude can influence high-level decisions during early clinical development. By providing a quick way to analyze complex data, the AI helps researchers prioritize which drug candidates have the best chance of reaching the market.

The path toward better models

Anthropic's strategy involves getting deeply involved in real-world drug programs to identify where their models succeed and where they fail. By working directly with biotech developers, the company can see how the AI handles specific scientific and administrative obstacles in real-time.

The current projects focus on a variety of conditions, including cancer, food allergies, and rare muscular disorders. While most of these treatments are still in experimental stages, the goal is to use AI to bridge the gap between molecular design and clinical trial preparation.

Moving forward, Anthropic intends to expand its involvement through more partnerships and specialized tools for life-science organizations. They aim to refine Claude's ability to navigate the nuances of scientific research, making it a more reliable partner for discovering lifesaving medicines.