Chris Gibson entered Dean Li’s laboratory at the University of Utah as an MD/PhD student planning to become a cardiothoracic surgeon. The group was studying cerebral cavernous malformation, or CCM, a rare condition in which abnormal blood vessels in the brain can leak and cause serious neurological harm.
The laboratory had accumulated a compelling mechanistic hypothesis. Its work pointed toward activation of a protein called RhoA as a central driver of the disease. If that model was right, inhibiting the pathway should improve the condition in an animal model. The test was designed to move the science forward.
Instead, the intervention made the disease worse.
There are many reasons an experiment can fail. The dose may be wrong. The model may be imperfect. The protocol may contain an error. A surprising result does not automatically disprove a theory. But Gibson has described the moment as humbling because it forced the team to confront how much it did not understand. The expert story of cause and effect had organized the work. The organism did not cooperate with the story.
Observation before explanation
The team changed the order of operations. Rather than beginning with a fully specified mechanism and testing compounds only against that account, it used images of diseased cells to observe what different compounds actually did. Could a compound make the cell’s visible state look healthier, even before the researchers knew exactly why?
This kind of phenotypic screening was not new. What mattered was the scale and system that the founders believed could be built around it. Advances in automated microscopy, computation, and data analysis made it possible to turn large numbers of cellular images into a search space. In 2013, Gibson, Li, and Blake Borgeson founded Recursion Pharmaceuticals around that approach.
The origin story can be overstated. One anomalous experiment did not, by itself, prove a new drug discovery platform. Recursion still had to develop disease models, generate reproducible data, secure funding, attract scientific and technical talent, and test whether cellular observations could yield medicines. The company’s later progress and setbacks remain separate evidence from the founding insight.
What the experiment changed was more fundamental than a single target. It changed the team’s posture toward knowledge. Mechanistic expertise remained valuable, but it no longer had to be the only entrance to discovery. Observation could reveal a useful relationship before the full explanation was available.
That is a difficult move for experts. Expertise creates pattern recognition, but it also makes a coherent interpretation feel complete. The Recursion story is powerful because the contradiction was not treated merely as noise or embarrassment. It became a reason to build a system that could remain curious at a scale no individual scientist could manage alone.
Sources and further reading
Keep exploring
Flatiron Health: When do clinical records become research evidence?Foundation Medicine: Turning a genome into a decision.Put the question to work with an AI prompt →Browse all cases →For a discussion tailored to your team, explore a workshop with Julien.