Can AI drug discovery transform cancer treatment outcomes?

Recursion Pharmaceuticals announced on Monday that its AI-discovered drug REC-4881 achieved measurable success in treating familial adenomatous polyposis (FAP), a rare genetic condition that causes hundreds of polyps in the colon and dramatically increases colorectal cancer risk. The results mark the first clinical validation of Recursion's AI platform after moving from laboratory discovery to human patients.

In the Phase 1b/2 TUPELO trial, 9 out of 11 patients maintained a durable reduction in total polyp burden, with a median reduction of 53% twelve weeks after stopping therapy. That durability matters. Previous treatments temporarily suppress polyps but require ongoing administration or surgery. This drug appears to create lasting remission.​

"This is the first clinical validation of Recursion's AI platform, where we garnered unbiased insights to know that this molecule could work in this disease and now proving it in patients," said Najat Khan, incoming CEO of Recursion.

The stock market agreed. Recursion shares rose 6% in premarket trading on the news. But the real significance is about what this result means for drug discovery, AI development, and a disease that currently offers only one real treatment. Surgical removal of the colon.

Familial adenomatous polyposis is a genetic disorder caused by mutations in the APC gene. People with FAP develop hundreds, sometimes thousands, of polyps throughout their colon, often by their teens or twenties. Without intervention, virtually all FAP patients develop colorectal cancer by age 40.​

For decades, the answer has been preventive colon surgery. Patients undergo total colectomy or proctocolectomy which is the surgical removal of the colon or colon and rectum. It prevents cancer but carries significant quality-of-life implications. Patients lose normal bowel function and require either an ileostomy (a pouch for waste) or an ileoanal pouch (a surgically created pouch that requires frequent trips to the bathroom).​

Some chemotherapy drugs like celecoxib and sulindac can suppress polyp growth, but they're not durable and don't address the underlying genetic mutation. They're Band-Aids on a genetic condition.​

REC-4881 targets biology differently. Rather than just suppressing polyp growth, it addresses how APC loss creates vulnerability. The drug inhibits MEK1/2 proteins, which Recursion's AI platform identified as a therapeutic vulnerability in APC-deficient cells.​

That insight is crucial. MEK1/2 inhibition hadn't previously been investigated as a treatment strategy for FAP in any clinical setting. Recursion's AI found a novel target, something human researchers hadn't conceived of.​

How Did AI Discover This?

REC-4881 was discovered using Recursion's "Recursion OS," an early version of their AI platform. The company ran high-content AI-driven phenotypic screening on APC-deficient human cell models. Rather than starting with published literature and known targets, the AI explored the entire biological landscape of FAP cells.​

The AI tested hundreds of compounds, millions of genetic relationships, and thousands of mechanistic possibilities. It found that APC-deficient cells had a specific vulnerability to MEK1/2 inhibition. It identified why the target worked, reverting diseased cells back toward a healthy-state phenotype.​

Then Recursion optimized small molecules to hit that target without hitting others. The result? REC-4881, a compound that's potent, selective, and concentration-dependent.​

This is the promise of AI drug discovery executed correctly. It's not automating what humans already do. It's asking questions humans never thought to ask, exploring search spaces humans can't navigate, and finding targets that would take traditional medicinal chemistry years to discover.

In the TUPELO trial, patients received REC-4881 at 4mg daily. After 12 weeks of treatment, 75% of evaluable patients showed reductions in total polyp burden​, median reduction was 43%​, and patients maintained this reduction for 12 weeks after stopping the drug​.

That last point is significant. The drug doesn't just suppress polyps while you're taking it. It appears to create lasting change in the disease biology.​

Side effects were generally mild to moderate, rash, diarrhea, elevated muscle enzymes (typical of MEK inhibitors). Only 16% experienced Grade 3 side effects. No Grade 4 or higher events occurred.​

Recursion is now expanding the trial to patients aged 18 and above (the initial cohort was age 55+) and will engage the FDA in the first half of 2026 to discuss a potential registration pathway for approval.​

Why Does This Matter?

FAP is rare, affecting roughly 1 in 10,000 people. But the broader implication is huge. Recursion just proved that AI-discovered drugs can reach human patients and deliver clinical benefit. This is the first real proof point that end-to-end AI drug discovery works.​

If Recursion can replicate this with other diseases, the entire model of drug discovery changes. Timelines compress. Costs drop. Novel targets that humans missed become treatable.

"We're one of the first to show these green shoots of proof points. It's really a good arc in terms of going from discovery AI to the clinic," Khan added.

Drug developers are increasingly adopting AI for discovery and safety testing. Experts suggest these technologies could cut development timelines and costs by at least 50% in the next few years while reducing animal testing.​

Recursion has additional programs in its pipeline. REC-1245, discovered through similar AI methodology, targets a novel cancer mechanism and is in Phase 1/2 trials. About 10 advanced discovery programs are progressing, with another expected to enter trials in 2025.