Recursion AI Drug Discovery: Timeline and Results

Recursion built its drug discovery engine on cell images and machine learning, then merged with Exscientia to add AI chemistry. Its clinical record is mixed, and this page lays out what has worked, what has not and how far each result has been proven.

Presented By Our Partners

The short answer: Recursion Pharmaceuticals uses automated cell imaging and machine learning to find drug targets and molecules, and it trains its own foundation models. Its most encouraging current signal is company-reported Phase 1b/2 data showing reduced polyp burden with REC-4881 in familial adenomatous polyposis. Its best-proven human result is still a peer-reviewed Phase 1 safety study, three clinical programs were deprioritized in 2025, and none of its drugs has been approved.

Tier: 2, AI science lab
Own models: Yes. OpenPhenom-S/16 for microscopy images (2024) and Boltz-2 for binding affinity, built with MIT (2025)
Own wet lab: Yes. Its Roche and Genentech deal announcement describes high-content screening at scale
First health move: December 2014, a peer-reviewed machine-learning screen of 2,100 known drugs for cerebral cavernous malformation by Recursion-affiliated scientists (PMID 25486933)
Furthest result so far: Tested in people, Phase 1 (REC-994, peer-reviewed). Its Phase 2 data for REC-994 and REC-4881 are company-reported
Tier 1 partners: None. We found no funding or partnership with OpenAI, Anthropic, Google DeepMind, Microsoft or Meta. Google Cloud distributes its OpenPhenom-S/16 model, and Nvidia, a chipmaker rather than a tier 1 lab, invested $50 million in 2023

What Is Recursion?

Recursion is a Salt Lake City drug discovery company that treats biology as a data problem: it photographs cells after genetic and chemical changes, then uses machine learning to link genes, diseases and compounds. It completed a merger with Exscientia, a UK AI chemistry company, in November 2024, according to Recursion’s announcement.

Its earliest dated health work we could source is a December 2014 Circulation paper by scientists listing Recursion as an affiliation. They ran an automated, machine-learning image screen of 2,100 known drugs in cells lacking the CCM2 gene, then tested hits in mice; vitamin D3 and tempol each cut lesion burden by about 50% in a mouse model of cerebral cavernous malformation (PMID 25486933). Tempol became REC-994 and was later tested in people (PMID 38655895). That origin matters for investors and doctors alike: Recursion’s early clinical drugs were known compounds found by AI-assisted screening, not molecules designed from scratch by AI.

Recursion builds models as well as drugs. It released OpenPhenom-S/16, a microscopy foundation model, in Google Cloud’s Model Garden in November 2024, and co-released Boltz-2 with MIT in June 2025. In November 2025 it announced a CEO transition plan (Recursion), with Najat Khan named to become CEO effective January 1, 2026, according to The Fly.

Recursion Healthcare Timeline: Every Move Since 2014

Entries are newest first. Result rows carry an evidence level from our fixed scale; deals, funding, hires and pipeline decisions are marked Not a result.

Date	Move	Type	Evidence level	Source
Dec 2025	Reported Phase 1b/2 TUPELO results for REC-4881, a MEK inhibitor, in familial adenomatous polyposis, describing reductions in polyp burden at 25 weeks.	Result	Company-reported, Phase 2 (Phase 1b/2 trial)	Recursion
Nov 2025	Announced a CEO transition plan, with Najat Khan to become CEO effective January 1, 2026.	Hire	Not a result	Recursion
Jun 2025	Released Boltz-2 with MIT, an open model that predicts protein structure and small-molecule binding affinity.	Model	Company-reported, Research tool	PMID 40667369
May 2025	Reworked its pipeline several months after the Exscientia merger, deprioritizing three clinical programs after a data review.	Trial	Not a result	Recursion via Wedbush
May 2025	Presented preliminary Phase 1b/2 TUPELO data for REC-4881 in familial adenomatous polyposis at Digestive Disease Week, showing reduced polyp burden.	Result	Company-reported, Phase 2 (Phase 1b/2 trial)	Recursion
Nov 2024	Completed its merger with Exscientia, a UK AI chemistry company.	Acquisition	Not a result	Recursion via Barchart
Nov 2024	Released OpenPhenom-S/16, a foundation model for microscopy images, in Google Cloud’s Model Garden.	Model	Company-reported, Research tool	Recursion via Nasdaq
Sep 2024	Reported Phase 2 data for REC-994 in cerebral cavernous malformation: the trial met its primary endpoint of safety and tolerability, but investors read the efficacy data as mixed.	Result	Company-reported, Phase 2	Recursion
Jun 2024	Published Phase 1 results for REC-994 (tempol) in 84 healthy volunteers: well tolerated at single and repeated doses up to 800 mg.	Result	Tested in people, Phase 1	PMID 38655895
Jul 2023	Nvidia invested $50 million and began a collaboration on foundation models for drug discovery.	Investment	Not a result	CNBC
Jan 2022	Exscientia, now part of Recursion, started a strategic research collaboration with Sanofi, reported as worth up to $5.2 billion.	Partnership	Not a result	Sanofi
Dec 2021	Collaboration with Roche and Genentech in neuroscience and oncology using machine learning and high-content screening at scale.	Partnership	Not a result	Recursion
Sep 2020	Collaboration with Bayer on fibrotic diseases, announced alongside a $239 million Series D led by Leaps by Bayer.	Partnership	Not a result	MedCity News
Dec 2014	Recursion-affiliated scientists used a machine-learning image screen of 2,100 known drugs to find vitamin D3 and tempol (later REC-994), each of which cut lesion burden by about 50% in a mouse model of cerebral cavernous malformation.	Result	Peer-reviewed, Preclinical	PMID 25486933

Recursion’s Biggest Results, and How Far Each Has Been Proven

Recursion has put several drugs into human trials and has had several setbacks. Here is each result with its evidence level, including the ones that did not hold up.

REC-4881 in familial adenomatous polyposis. Company-reported, Phase 2 (Phase 1b/2 trial). REC-4881 is a MEK inhibitor, according to the NCI Drug Dictionary. Recursion presented preliminary TUPELO data at Digestive Disease Week in May 2025 showing reduced polyp burden (Recursion, PharmaShots), then, at a readout scheduled for December 8, 2025 (Recursion via Barchart), reported reductions at 25 weeks that it called rapid and durable (Recursion). We found no peer-reviewed paper on these data in PubMed. What would prove it next: a peer-reviewed publication and a randomized, registrational trial.

REC-994 in cerebral cavernous malformation. Tested in people, Phase 1 (peer-reviewed), then a company-reported Phase 2. The Phase 1 studies in 32 single-dose and 52 multiple-dose healthy volunteers found REC-994 well tolerated up to 800 mg with no dose-limiting adverse effects (PMID 38655895). In September 2024 Recursion reported that a Phase 2 study met its primary endpoint of safety and tolerability (Recursion), but shares fell (Benzinga) and the efficacy data were widely read as mixed (Seeking Alpha). The program’s assets were later transferred to the patient group Alliance to Cure. REC-994 is tempol, a known compound found by screening, so it tested Recursion’s screening platform rather than AI drug design. It cleared safety hurdles but did not become a Recursion product.

Featured Partner

Invest in the Infrastructure Behind Modern Medicine

As healthcare expands beyond hospital walls, the buildings and campuses supporting that shift are generating compelling returns for investors who move early. The Healthcare Real Estate Fund offers qualified investors direct access to a curated portfolio of medical office, outpatient, and specialty care facilities.

Learn More →

Pipeline cuts after the merger. Not a result, but part of the record. In its first quarter 2025 business update in May 2025 (Recursion), several months after the Exscientia merger, Recursion reworked its pipeline (Fierce Biotech) and deprioritized three clinical programs following a data-driven review (Nasdaq). Cutting programs is normal in drug development, but it means the platform’s clinical yield is still unproven.

Boltz-2 binding affinity model. Company-reported, Research tool. The Boltz-2 preprint, by MIT and Recursion authors, reports that it is, to the authors’ knowledge, the first AI model to approach the accuracy of free-energy perturbation methods for small molecule binding affinity while being at least 1,000 times more computationally efficient, and it was released with open weights and code (PMID 40667369, Bio-IT World). The preprint is not peer reviewed. Because the model is open, outside groups can test it, which makes independent replication realistic. What would prove it next: peer review and independent benchmarks.

Phenomics and OpenPhenom. Company-reported, Research tool. Recursion’s microscopy foundation model OpenPhenom-S/16 is available through Google Cloud (Recursion). We found no peer-reviewed evaluation of it in PubMed. What would prove it next: a drug candidate credited to the phenomics maps succeeding in a randomized trial.

Who Recursion Works With

Recursion’s main pharma partners are Bayer, Roche and Genentech, and Sanofi. Bayer signed a fibrotic disease collaboration in September 2020, alongside a $239 million Series D led by Leaps by Bayer (Global Genes, MedCity News). Roche and Genentech signed a neuroscience and oncology collaboration in December 2021 built on machine learning and high-content screening at scale (Recursion). Sanofi came through Exscientia, which began a strategic research collaboration with Sanofi in January 2022 (Sanofi) that Pharma Manufacturing reported as worth up to $5.2 billion. Headline values like that are mostly milestone payments that depend on future success.

Tier 1 ties: none. We found no funding or partnership with OpenAI, Anthropic, Google DeepMind, Microsoft or Meta. Google Cloud hosts its OpenPhenom-S/16 model, a distribution arrangement rather than funding. Nvidia, a chipmaker rather than a tier 1 lab, invested $50 million in July 2023 and collaborates with Recursion on foundation models (CNBC). Recursion runs its own screening labs, and its model work with MIT on Boltz-2 is open source.

What to Watch Next

These are the measurable events that would move Recursion’s evidence level.

A peer-reviewed publication of the REC-4881 TUPELO results and any plan for a registrational trial in familial adenomatous polyposis.
Option exercises, milestone payments or named candidates disclosed under the Roche and Genentech, Sanofi or Bayer collaborations.
Peer review of Boltz-2 and independent benchmarks of its binding affinity predictions.
Pipeline additions or cuts in Recursion’s quarterly reports under its new CEO.
Any new Phase 2 readout from the combined Recursion and Exscientia pipeline.

Related: see how every lab is graded in our AI healthcare breakthroughs tracker, where AI-designed drugs stand in our status report on AI-designed drugs in human trials, the competing approach at Insilico Medicine, the layers of the AI drug discovery stack and how cell-level models work in virtual cell AI.

What This Means for You

If you are a patient, no Recursion drug is approved. People with familial adenomatous polyposis may want to ask their doctor about REC-4881 trials listed on ClinicalTrials.gov, but the current data are early and company-reported, and preventive surgery and surveillance remain the standard of care.

If you work in research or care, Recursion’s record is a useful check on AI drug discovery: its platform has put several known and new compounds into people, but Phase 2 is where the hard test happens, and some programs did not hold up. Its open Boltz-2 model is something your team can evaluate directly rather than taking any report on trust.

Compiled by Healthcare Discovery from company announcements, named news reports and PubMed. Every PMID on this page was checked against PubMed on October 28, 2026. Evidence levels follow the scale explained in our AI healthcare breakthroughs tracker. Healthcare Discovery uses AI research tools from several companies on this page, including OpenAI, Anthropic and Google models; no company listed reviewed or paid for this page.

What does Recursion Pharmaceuticals do?

Recursion is a Salt Lake City drug discovery company that images cells under many genetic and chemical changes and uses machine learning to find drug targets and molecules. It trains its own models, including the OpenPhenom microscopy model, and co-developed the Boltz-2 binding model with MIT.

Did Recursion merge with Exscientia?

Yes. Recursion completed its merger with Exscientia, a UK company focused on AI-driven chemistry, in November 2024. In May 2025 it reworked the combined pipeline and deprioritized three clinical programs.

Has Recursion had clinical trial setbacks?

Yes. Its REC-994 program for cerebral cavernous malformation met the primary safety and tolerability endpoint of a Phase 2 study in 2024, but the efficacy data were read as mixed and the assets were later transferred to the patient group Alliance to Cure. In May 2025 Recursion deprioritized three clinical programs after a data review. Its most positive current signal, REC-4881 for familial adenomatous polyposis, comes from company-reported Phase 1b/2 data.

What is Boltz-2?

Boltz-2 is an open AI model released by MIT and Recursion in June 2025 that predicts protein structure and how strongly small molecules bind to proteins. Its preprint reports performance approaching physics-based free energy methods while running at least 1,000 times more efficiently. It has not yet been peer reviewed.

Does Nvidia invest in Recursion?

Yes. Nvidia invested $50 million in Recursion in July 2023 as part of a collaboration on foundation models for drug discovery. Recursion also distributes its OpenPhenom-S/16 model through Google Cloud’s Model Garden.

Has a Recursion drug been approved?

No. As of October 2026 none of Recursion’s drugs has been approved. Its most advanced positive data come from a company-reported Phase 1b/2 trial of REC-4881, and its only peer-reviewed human data are Phase 1 safety results for REC-994.

Sources

Gibson CC, et al. Strategy for identifying repurposed drugs for the treatment of cerebral cavernous malformation. Circulation, 2014. PMID 25486933
Alfa R, et al. Clinical pharmacology and tolerability of REC-994, a redox-cycling nitroxide compound, in randomized phase 1 dose-finding studies. Pharmacology Research and Perspectives, 2024. PMID 38655895
Passaro S, et al. Boltz-2: Towards Accurate and Efficient Binding Affinity Prediction. bioRxiv (preprint), 2025. PMID 40667369
Recursion. Positive Phase 1b/2 results from ongoing REC-4881 TUPELO trial demonstrate rapid and durable reductions in polyp burden in familial adenomatous polyposis (FAP) at 25 weeks, December 2025. Recursion
Recursion via Barchart. Recursion announces webinar for upcoming clinical data readout on the TUPELO Phase 1b/2 trial of REC-4881 in familial adenomatous polyposis on December 8, 2025. Barchart
Recursion. Preliminary Phase 1b/2 Data for REC-4881 in Familial Adenomatous Polyposis (FAP) Demonstrates Reduced Polyp Burden, May 2025. Recursion
PharmaShots. Recursion presents preliminary Phase Ib/II TUPELO trial data of REC-4881 for FAP at DDW 2025, May 2025. PharmaShots
National Cancer Institute. NCI Drug Dictionary: MEK inhibitor REC-4881. National Cancer Institute
Recursion. Recursion Announces CEO Transition Plan to Drive Next Phase of Growth, November 2025. Recursion
The Fly via TipRanks. Recursion appoints Najat Khan as CEO effective January 1, November 2025. TipRanks
Bio-IT World. MIT Researchers Unveil Boltz-2: AI Model Predicts Protein Structure, Binding Affinity in Seconds, June 2025. Bio-IT World
Fierce Biotech. Several months after Exscientia merger, AI biotech outfit Recursion reworks pipeline, 2025. Fierce Biotech
Nasdaq. Recursion deprioritizes 3 clinical programs following data-driven review, stock down. Nasdaq
Recursion via Wedbush. Recursion Reports First Quarter 2025 Financial Results and Provides Business Update, May 2025. Wedbush
Recursion via Barchart. Recursion Completes Merger with Exscientia, Launches Comprehensive AI-Enabled Drug Discovery Portfolio, November 2024. Barchart
Recursion via Nasdaq. Recursion Announces the Release of OpenPhenom-S/16 in Google Cloud’s Model Garden, November 2024. Nasdaq
Recursion. Recursion Announces Phase 2 Data of REC-994, a First-in-Disease Investigational Treatment, September 2024. Recursion
Benzinga. Recursion Pharmaceuticals Stock Is Sliding Tuesday: What’s Going On?, September 2024. Benzinga
Seeking Alpha. Nvidia-backed Recursion hurts AI biotechs on mixed data for lead drug. Seeking Alpha
Alliance to Cure. REC-994 Assets Transferred. Alliance to Cure
CNBC. Nvidia invests in biotech company Recursion for AI drug discovery, July 2023. CNBC
Sanofi. Press release on its research collaboration with Exscientia, January 7, 2022. Sanofi
Pharma Manufacturing. Exscientia, Sanofi ink potential $5.2B AI research deal, January 2022. Pharma Manufacturing
Recursion. Recursion Announces Transformational Collaboration with Roche and Genentech in Neuroscience and Oncology, December 2021. Recursion
MedCity News. Recursion raises $239M in Series D, partners with lead investor Bayer, September 2020. MedCity News
Global Genes. Recursion Raises $239 Million and Strikes Fibrotic Diseases Collaboration with Bayer, September 2020. Global Genes

Free Daily Briefing

The Latest Longevity Science.
Delivered Every Morning.

Join researchers, physicians, and health professionals getting daily breakthroughs in AI-driven medicine, epigenetics, and longevity research.

Support the research that powers this editorial

No spam. Unsubscribe anytime. We respect your inbox.