Anthropic Acquires Mithril Medicine to Test AI-Supported Model for Rare Disease Drug Development

Rare Daily Staff

Anthropic, the artificial intelligence company behind Claude, has acquired Mithril Medicine, bringing two early-stage muscular dystrophy research programs in-house as it tests whether a single scientist working with its AI model can direct multiple drug development programs.

The acquisition, completed in July, makes Mithril’s work on LAMA2-related congenital muscular dystrophy and alpha-dystroglycanopathy Anthropic’s first internal drug development programs. The company discussed the acquisition on its website in an update on its work in the life sciences. Financial terms were not disclosed.

Before the acquisition, Mithril had one full-time employee—founder Cody Tranbarger—supported by part-time consultants and outside laboratories. Tranbarger used Claude to help design and plan experiments, analyze data, and coordinate research across 11 external labs.

Mithril’s two programs target diseases that weaken the connection between muscle fibers and the supportive protein layer surrounding them. Without that connection, movement can damage muscle fibers, which are progressively replaced by fat and scar tissue.

The company is developing antibodies designed to act as bridges, reconnecting muscle fibers to that support layer. These antibodies are called bispecific because they attach to two different targets: one on the muscle fiber and another in the surrounding layer. The goal is to restore the connection and protect the fibers from damage.

Anthropic said Mithril demonstrated the potential of its antibody approach to LAMA2-related congenital muscular dystrophy, or LAMA2-CMD, in less than 18 months. In a short-term study, treated mice with a severe form of the disease had roughly twice the grip strength and triceps muscle force of untreated animals, and more survived to the study’s end.

The treated mice remained weaker than healthy animals. The findings do not establish that the approach is safe or effective in people.

Mithril has not yet selected formal drug candidates. The company aims to do so in 2027, with clinical trials potentially beginning in late 2028 for LAMA2-CMD and in 2029 for alpha-dystroglycanopathy. Those dates are targets, not commitments.

Anthropic describes Mithril as its first experiment with a “Pipeline-in-a-Person” research model: a scientist sets the direction and makes decisions while Claude helps maintain program records, coordinate outside laboratories, and interpret their results.

The premise is that reducing administrative work could allow scientists to pursue more research programs simultaneously, potentially making drug development more practical for rare diseases with small patient populations.

According to Anthropic, coordinating vendors and tracking documents, data, and decisions had become a bottleneck at Mithril. Claude helped Tranbarger reconstruct the status of experiments from email threads and 14 months of slide presentations when responsibility for the work changed hands. It identified obstacles and drafted five handover emails, which Tranbarger reviewed and sent to two outside laboratories.

In another example, Claude consolidated hundreds of antibody-screening files into a list of promising antibodies and assessed how readily they might be developed into drugs.

Tranbarger estimated that turning an initial idea for a new program into a detailed research plan and experimental strategy took two to three days with Claude, rather than the one to two months he believed it would otherwise require. Those figures are his estimates, not results from a controlled comparison.

Tranbarger continues to direct Mithril’s two original programs and is preparing to launch several more focused on rare pediatric diseases without approved therapies. Under the model, he retains responsibility for choosing experiments, deciding whether programs advance or stop, and approving work.