Anthropic publishes early progress on Claude in drug development
Anthropic published an account of Claude's progress in drug development on Oct 27, detailing results from work with six biopharma partners and internal programs. Six scientists at Generate Biomedicines used Claude Code to propose CAR-T designs for solid tumors, with one design shortening the genetic instruction portion by about 70 percent, immune cells carrying roughly ten times as many receptors, and better tumor killing and cell longevity in the lab. At Divergence Labs, a two-arm molecule program for food allergy had the model generate roughly 35,000 antibody designs largely autonomously in about a month, finding three designs that bind IgE and 20 that dislodge it, with three of the latter outperforming the standard medication, and the next step being to combine the two arms. At BigHat Biosciences, model-proposed changes delivered through the Reccy lab data system produced about 1,010 variants designed and tested in roughly ten weeks, with the best versions binding over ten times more tightly, the cancer-targeting lead performing 18 degrees better in solution, and the longer-acting lead clumping over a hundred times more slowly. At Enveda, model review of 1,061 pages of manufacturing documents caught all 11 issues human reviewers found plus seven they missed, cutting review time from over 25 hours to under 3 hours, an 89 percent reduction, and one regulatory first draft took 4 hours of model time against about three months manually. The Column Group's new Pleiades company evaluated 45 target-disease pairs with its first skill, and Bristol Myers Squibb used the model to assess early trial plans and probability of success, planning to expand in 2027. Anthropic also disclosed acquiring Mithril Medicine, whose founder used Claude to coordinate 11 external labs and advance a rare muscular dystrophy therapy to promising mouse results in under 18 months, potentially reaching human trials as soon as late 2028. The article stresses the results are early and untested in humans.