Recursion Pharmaceuticals published data on December 8, 2025 showing that REC-4881, a drug its AI platform selected through phenotypic screening rather than through a hypothesis-driven target search, cut polyp burden by a median of 43% after 13 weeks in patients with familial adenomatous polyposis, a genetic condition that currently has no approved pharmacologic treatment anywhere. By week 25, after a treatment break, that reduction deepened to a median of 53%, and the company says it is now working toward a conversation with the FDA about what a registrational trial for the drug could look like.
A Disease Where Surgery Is Still the Only Real Option
Familial adenomatous polyposis, or FAP, is caused by inherited loss-of-function mutations in the APC tumor suppressor gene, and it causes patients to develop hundreds to thousands of adenomas throughout the colon and duodenum starting often in adolescence or young adulthood. Left unmanaged, the condition carries close to a 100% lifetime risk of colorectal cancer, and the current standard of care is a combination of intensive endoscopic surveillance and prophylactic colectomy, meaning removal of the colon before cancer has a chance to develop, typically at a young age. No drug has ever been approved specifically to treat FAP, leaving patients and their families with surgery and surveillance as the only tools available for a disease that is entirely predictable from a genetic test yet has no medical way to intervene before it progresses.
How the AI Found the Drug
Recursion did not set out looking for a MEK1/2 inhibitor. Instead, the company ran unbiased phenotypic screens across thousands of compounds inside human cell models engineered to be deficient in the APC gene, using high-content imaging paired with machine-learning models to look for any compound that visibly reverted the diseased cellular phenotype back toward something resembling healthy tissue, rather than testing compounds against a single predetermined biological target. REC-4881 emerged from that screen as what the company describes internally as a rescue hit, a compound whose effect on the diseased cells was strong enough to pursue into human trials, arrived at through a discovery process that looked nothing like traditional target-first drug design.
Inside the Trial Data
The results come from TUPELO, an open-label Phase 1b/2 study that initially enrolled patients aged 55 and older, with the company now working to expand eligibility down to age 18. In the Phase 2 cohort, 12 patients were efficacy-evaluable at the 4 milligram once-daily dose, with 75% showing any reduction in polyp burden by week 13 and a median reduction of 43%. By week 25, among 11 evaluable patients, 82% showed a reduction, with 73% of patients achieving at least a 30% decrease in polyp count, and four of ten patients showed at least a one-point improvement on the Spigelman staging system used to track FAP severity, an improvement that persisted through week 25. Treatment-related side effects occurred in nearly all patients, mostly a dermatitis-like rash and elevated creatine phosphokinase levels, with grade 3 events in roughly 16% of patients, no grade 4 or higher events, and two patients requiring temporary dose interruptions.
Recursion’s Own Framing of the Milestone
Recursion CEO Najat Khan has described this data as a turning point for the company’s broader strategy, saying Recursion has reached an inflection point moving from proving that AI can participate in drug discovery to demonstrating that what she calls an AI-native operating system can generate actual clinical proof and durable value. Khan tied the FAP results to a separate milestone reached with partner Sanofi, calling both signs of what she described as the increasing maturity of a deeply integrated model. Underscoring how central data infrastructure is to that model, Recursion says it built a large language model trained on more than 250,000 patient records specifically to establish a natural history baseline for FAP, giving the company a comparison point for what an untreated patient’s disease trajectory typically looks like.
What Skeptics Will Want to See
A median 43-to-53% reduction in polyp burden from an open-label study with fewer than a dozen efficacy-evaluable patients at the key dose is a meaningfully different evidentiary bar than a randomized, placebo-controlled trial with hundreds of participants, and Recursion has not yet run the larger study that would satisfy that higher bar. The adverse event profile, with treatment-related side effects in nearly all patients and grade 3 events in roughly one in six, also means dosing and tolerability will be scrutinized closely as the program moves toward any larger trial, particularly if the FDA asks for data in a broader age range once eligibility expands to adults as young as 18. FAP is also a slow-moving, chronic condition, so a 25-week readout, however promising, does not yet answer how durable any polyp reduction is over the years that actually matter for a patient’s colorectal cancer risk.
What Happens Next
Recursion has already secured Fast Track and Orphan Drug designations from the FDA along with orphan status in Europe, regulatory tools designed to speed development for exactly the kind of rare, high-unmet-need disease FAP represents, and the company has a meeting planned with the FDA to discuss registration strategy and optimal dosing. If that conversation produces a clear path to a pivotal trial, REC-4881 would be positioned to become the first ever approved pharmacologic treatment for FAP, a milestone that would matter both for patients currently facing a choice between intensive surveillance and preventive surgery, and for Recursion’s argument that its AI-native discovery platform can find real drugs in diseases too rare or too complex for conventional target-based approaches to prioritize.
