Anthropic disclosed on 27 October that it bought Mithril Medicine, which had one full-time employee, in July, after acquiring Coefficient Bio, which had fewer than 10 staff, R&D World reported. The Cambridge, Massachusetts-based startup’s two rare-disease programmes became Anthropic’s first internal drug-development programmes. Anthropic also detailed work with six organisations using Claude in drug development, Crypto Briefing reported.
Key points
- Mithril founder Cody Tranbarger used Claude to help coordinate work across 11 outside laboratories.
- The six collaborators include Bristol Myers Squibb, BigHat Biosciences and Generate Biomedicines.
- Anthropic says Mithril reached proof of principle in mice for one muscular dystrophy programme in under 18 months.
- Mithril aims to choose formal drug candidates in 2027, with human trials for its lead programme possible as early as late 2028.
Mithril coordinated work across 11 laboratories
Founder Cody Tranbarger directed part-time scientific consultants and 11 contract laboratories, using Claude to help organise experiments, data and logistics. Mithril’s team planned and supervised the research. The outside laboratories made antibodies, tested their binding to targets, ran cell assays and carried out mouse studies.
Eight paid consultants and advisers supplied expertise in biology, antibody engineering and research planning. Tranbarger graduated from Harvard’s joint MS/MBA biotechnology programme in 2025. He had worked at gene-editing company Prime Medicine and life-sciences investor Casdin Capital before becoming an entrepreneur-in-residence at biotech venture firm Atlas Venture.
The arrangement extends Anthropic’s role from supplying Claude to research teams into owning drug programmes. Coefficient Bio, its earlier acquisition, was a biology-focused AI startup with two cofounders. Another AI company has pursued an acquisition in the field: AI Affairs reported OpenAI’s purchase of Orchestra Bio to add drug-development agents to Rosalind.
Bristol Myers Squibb joins six-company roster
Anthropic’s collaborators are BigHat Biosciences, Bristol Myers Squibb, Divergence Labs, Enveda, Generate Biomedicines and The Column Group. Their work uses Claude and its Claude Science workbench across tasks ranging from early searches for drug candidates to regulatory documentation. Most of the projects remain preclinical.
Bristol Myers Squibb has used Claude since May 2026 to plan clinical trials, including assessments of their likelihood of success. Generate Biomedicines reported that Claude helped produce genetic instructions for a CAR-T therapy that were approximately 70% shorter. CAR-T therapies modify a patient’s immune cells to attack disease.
BigHat Biosciences said its use of Anthropic’s technology produced more than 10x tighter binding, a measure of how strongly a candidate attaches to its target, as well as improvements in stability. Divergence Labs used Claude to generate approximately 35,000 antibody designs for food allergies. Enveda reported an 89% reduction in time spent on regulatory and documentation processes. Life-sciences investor The Column Group used Claude for operational work while setting up a biotech company called Pleiades.
The partner work covers different points in development. Bristol Myers Squibb’s use concerns trial planning, while Enveda’s reported result concerns documentation time. Anthropic also has internal programmes aimed at two rare muscular dystrophies through Mithril.
Mithril’s LAMA2 study tested seven mice
Mithril’s lead programme addresses LAMA2-related congenital muscular dystrophy. Changes in the LAMA2 gene weaken the connections between muscle fibres and surrounding tissue. Mithril developed antibodies designed to bind two targets and bridge that connection.
Anthropic says Mithril reached proof of principle in mice in under 18 months. In a technical report awaiting peer review, seven antibody-treated mice recorded about twice the grip strength and maximum upper-arm muscle strength of four mice dosed with a control liquid, R&D World reported. The molecules tested were early research compounds rather than selected drug candidates.
The researchers assigned treatment by litter without randomisation. They tested the treated group 126 to 161 days after the controls, and moved the mouse colony between the tests. Those differences complicate attributing the measured gains to the treatment. The report calls for a repeat study with concurrent, blinded testing.
Claude generated a 341-page production report covering 113 antibody constructs in under eight minutes during a week the team chose as its best recent example. After a team member rejected one recommendation, Claude revised the research plan and related drafts to cover 89 constructs within four minutes. At least three of six Claude errors examined by the team reached outside laboratories in emails edited and sent by people, R&D World reported.
Mithril aims to select formal drug candidates in 2027. If further development succeeds, its LAMA2 programme could enter human trials as early as late 2028. The company targets trials for its second programme, focused on alpha-dystroglycanopathy, as early as 2029.
