In-house experiments could offer an edge in drug development as AI becomes widely available, suggests physician-scientist David Shaywitz

In-house experiments could offer an edge in drug development as AI becomes widely available, suggests physician-scientist David Shaywitz

On October 30, biotech entrepreneur Daphne Zohar asked: what sets a company that calls itself an AI drug developer apart from its competitors when models are available to everyone and changing rapidly? In response, Shaywitz suggested that an advantage could come from understanding what causes disease by combining data on how it manifests with results from in-house experiments.

Zohar co-founded the biotech company Karuna Therapeutics. In March 2024, pharmaceutical company Bristol Myers Squibb acquired all its shares for approximately 14 billion dollars. Zohar herself expects AI to have an enormous impact on the industry.

Zohar cited an essay by biotech investor Bruce Booth published that day. Booth predicts that over the next three to five years, faster molecular design will intensify competition around known targets, such as proteins that a drug is intended to act on. Many models are trained on data about known molecules and biological mechanisms, making it easier for more teams to develop their own drug candidates against the same targets.

In Booth’s scenario, companies compete for contract research and manufacturing providers, trial sites, and eligible patients. Competition for patients slows trial recruitment; comparing new drugs with existing treatments takes years of work and investment. Booth expects overall financial returns to fall: many teams incur costs, while only some drugs will demonstrate an advantage over existing treatments. Individual successes may improve treatment for patients.

Shaywitz proposed generating proprietary data on the causes of disease by studying its manifestations and experimentally testing which interventions alter disease-related processes. He suggests that this work could yield “a proprietary understanding of causal relationships in biology” and help companies select new targets with greater confidence.

In August, insitro founder Daphne Koller distinguished three tasks in drug development: identifying the disease mechanism, creating a molecule, and testing the treatment in humans. Shaywitz proposes seeking an advantage at the first step. He cited insitro as one example of this approach.

According to insitro’s description, human genetic and health data help researchers select cells for experiments and decide what to measure in them. For example, researchers compare fat accumulation in the human liver with fat accumulation in cultured liver cells. By altering gene activity in these cells, they test how each gene affects fat accumulation. AI compares the experimental results with human genetic and health data to propose a drug target. The experimental results are fed back into the models to help select subsequent experiments.

Zohar agreed on the importance of biology and clinical work and clarified: “My question is whether AI provides a competitive advantage or helps everyone in the market.”