AI will help small teams evaluate more ideas for new drugs, investor Simon Barnett predicts
AI will help small teams evaluate more ideas for new drugs, investor Simon Barnett predicts
On October 30, Simon Barnett, a partner at the investment firm Dimension, proposed using AI to gather information and coordinate work with external laboratories. In a broadcast by the pharmaceutical news outlet Endpoints News, journalists Drew Armstrong and Andrew Dunn asked him how this changes research and who pays for the experiments.
The discussion followed AI developer Anthropic’s announcement that it had acquired Mithril Medicine, a biotechnology company with one full-time employee. At Mithril, the AI assistant Claude drafted emails to laboratories, and a human edited and sent them. Pharmaceutical companies have long entrusted development to separate companies set up for individual projects: in 2011, Eli Lilly transferred the development rights for a migraine drug candidate to Arteaus, then bought them back in 2014 after trials in patients. When a small company uses contractors, it pays for the experiments, while staff at external laboratories carry them out.
Barnett proposes bringing an experienced drug developer and an AI specialist together in a small team. AI would gather information about competitors, run specialized software to calculate protein structures, and coordinate work with external laboratories. Barnett cites voice AI assistants that call these laboratories as an example. Payroll savings, he says, “do not determine whether this model succeeds or fails.”
Barnett argues that spending less time on coordination would allow the team to evaluate more related ideas in parallel during early development: for example, several ways to activate cell receptors, which are proteins that receive signals. The drug developer assesses whether experiments in cells or animals are suitable for testing each idea, what the chosen laboratory assay can reveal, and how to test the drug in humans later. The AI specialist decides which tools to use and what data the team still needs to collect. This helps the team decide which ideas to discard and which experiments to fund.
Barnett sees a commercial niche for these companies: a project may appeal to a small team even if its market is too small to interest a large pharmaceutical company.
In an essay discussed during the conversation, investor Bruce Booth predicts that AI will make molecules easier to create and increase the number of drugs in development that act on already popular targets, including proteins and other entities in the body. Developers of similar drugs will compete for laboratories, trial sites, and patients. Competition for patients will prolong trial recruitment. Barnett expects projects that struggle to distinguish themselves to have more difficulty attracting investment.
Experiments require laboratory animals, and human trials require volunteers. Asked whether there will be enough animals and trial participants for the growing number of projects, Barnett replies: “We have a limited number of monkeys.”