Recursion Pharmaceuticals (PRNewsfoto/Recursion Pharmaceuticals)
Recursion Pharmaceuticals dropped phase 1b/2 results Monday for REC-4881, a MEK inhibitor it flagged through AI screens. In 12 older patients (55 years old and older) with familial adenomatous polyposis, the drug trimmed total polyp burden a median 43% after three months. The gains stuck around: 12 weeks off treatment, the median drop climbed to 53% in 11 patients.
FAP stems from pathogenic loss-of-function mutations in the APC tumor suppressor gene. The syndrome drives the growth of hundreds, sometimes thousands, of adenomas in the colon and duodenum, and the lifetime risk of colorectal cancer approaches 100% without intensive surveillance and prophylactic colectomy. Despite decades of interest in chemoprevention, there are still no approved pharmacologic therapies. Most patients cycle through frequent endoscopic polypectomies and major abdominal surgery in early adulthood.
Fast signal, early days
TUPELO is an open-label phase 1b/2 study testing REC-4881 as monotherapy in adults with classical FAP and measurable polyps in the duodenum or residual colorectum. The trial initially restricted enrollment to patients 55 years and older, a small and older cohort that limits generalizability to the broader FAP population, which skews younger and often undergoes colectomy in the teens or twenties. Recursion plans to expand eligibility to patients 18 and older.
Efficacy is assessed endoscopically at baseline, week 13 on treatment, and week 25 after a 12-week washout period, using the summed diameter of all visible polyps as the primary endpoint.
In the phase 2 cohort, 12 efficacy-evaluable patients who received at least 75% of planned dosing with REC-4881 at 4 mg once daily had a median 43% reduction in total polyp burden at week 13. Seventy-five percent of patients had some degree of reduction. At week 25, 11 patients were evaluable and the median reduction deepened to 53%, with 82% showing any reduction and 73% achieving at least a 30% decrease in summed polyp diameter compared with baseline. Four of 10 patients had at least a 1-point improvement in Spigelman stage, a scoring system for upper gastrointestinal polyposis that helps guide surveillance and surgical decision-making, and these improvements persisted at week 25.
Recursion positions these results against the natural history of FAP. A registry analysis from Amsterdam and a United States electronic health record study both suggest that most untreated patients experience year-over-year increases in polyp burden, with spontaneous regression rare.
The current TUPELO dataset remains small and single-arm, so the findings are best viewed as hypothesis-generating rather than definitive.
Safety in line with MEK class
Safety data from 19 patients across the phase 1b and phase 2 cohorts are preliminary but broadly consistent with MEK inhibitor class effects. Nearly all patients reported treatment-related adverse events, most commonly dermatitis-like rash and creatine phosphokinase elevations, with most events graded 1 or 2 in severity. Grade 3 treatment-related events occurred in about 16% of participants, and no grade 4 or higher treatment-related events have been reported so far. Two patients required dose interruptions.
Longer-term safety and tolerability will be a key focus if REC-4881 moves into larger studies in younger patients, where preventive treatment could mean years of exposure.
Phenotypic AI in the spotlight
REC-4881 is also a platform story. Recursion identified MEK1/2 inhibition as a potential way to counter APC loss of function by running an unbiased phenotypic screen of thousands of compounds in APC-deficient human cell models. The company used high-content imaging and machine-learning models to extract high-dimensional features that distinguish diseased from healthy cell states and then scored compounds on their ability to revert the diseased phenotype.
REC-4881 emerged as one of the strongest “rescue” hits, shifting APC-deficient cells toward a healthy-like phenotype and dampening hyperactive ERK/MAPK signaling downstream of APC loss. On the back of those data, Recursion in-licensed the molecule from Takeda, where it had been studied in solid tumors, and redirected it into FAP. Company leaders are now highlighting TUPELO as the first full “validation cycle” of the Recursion OS, linking an image-based phenotypic signal to a mechanism, a clinical asset and human data in a disease with no approved drugs.
How it compares with other non-surgical approaches
Recursion is not alone in seeking a pharmacologic alternative to FAP’s surgical default. Biodexa Pharmaceuticals has reported 12-month phase 2 results for eRapa, an oral rapamycin formulation, showing a median 17% reduction in overall polyp burden and a 75% non-progression rate across all patients, with one cohort achieving a 29% median reduction and close to 90% non-progression. That program has secured Fast Track designation.
Direct comparisons between REC-4881 and eRapa are not possible, given differences in study design, patient mix, follow-up duration and endpoints. eRapa has longer 12-month data in a larger cohort. REC-4881 shows a larger median reduction over a much shorter window plus an off-treatment durability signal. How regulators and clinicians weigh those trade-offs will depend on upcoming randomized or controlled datasets, long-term safety, and how workable these regimens are for largely young adult patients who may otherwise be facing major surgery.
What comes next
REC-4881 has Fast Track and Orphan Drug designations in the United States and Orphan status in Europe. Recursion plans to expand TUPELO eligibility down to age 18 and to meet with the FDA in the first half of 2026 to discuss a potential registration strategy and optimized dosing schedules. Those plans are subject to regulatory feedback and to how the emerging data look as more patients and longer follow-up are added.
For now, TUPELO strengthens the case that non-surgical strategies for FAP are finally moving out of the purely theoretical. It also raises questions that extend beyond this rare syndrome. Can an image-based phenotypic platform repeatedly produce candidates with clinically meaningful effects in genetically defined diseases, or is REC-4881 an early, biology-favored exception. How much evidence of cancer risk reduction regulators will ultimately require in a prevention setting remains an open question, as does the level of endoscopic polyp reduction that would be enough to shift a standard of care built for decades around the operating room.
Filed Under: machine learning and AI
