NEXT-GEN BIOTECH OS: v4.8 ARCHITECTURE LIVE

Industrializing
Drug Discovery
Through Code & Cells.

Decoding disease biology at industrial scale. By fusing automated high-throughput wet labs with billions of multimodal cellular parameters, we transform serendipitous discovery into predictable computation.

Explore Active Pipeline ↗ Platform Capabilities 🔬
REC-FEED LIVE POD 12 // 2.2M EXTRACTS/WK
⚡ Automated High-Content Phenomics 99.98% Latency Target
Neural Inference Rate 430 TFLOPS/BioPlate
Wet-Lab Throughput 24/7 Fully Autonomous
💾 Phenomic Data
25+ PB
Proprietary biological image & multi-omic datastore
🤖 Wet Lab Execution
2.2M+
Robotic microplate assays synthesized weekly
💊 Therapeutic Programs
10+ Clinical
Oncology, rare diseases, & neuroscience candidates
🧠 Generative Models
BioHive-2
Top-500 supercomputing cluster trained on human biology
⚙️ THE RECURSION OS ARCHITECTURE

A Closed-Loop System Decoding Human Cellular Biology

The Recursion Operating System replaces serendipitous discovery with an industrial flywheel. Automated robotic wet labs generate petabytes of high-content biological data to train our BioHive supercomputers, continuously refining predictive in silico models.

Continuous Biological Feedback Loop

Autonomous Discovery Cycle

CYCLE LATENCY: < 72 HOURS PER ITERATION
🤖 PHASE 01

Automated Wet Labs

Robotic liquid handlers run up to 2.2M microscopic cell assays weekly with gene knockouts and chemical perturbations.

High-Content Imaging ➔
💾 PHASE 02

Phenomic Datastore

Standardized image acquisition captures cellular morphology and transcriptomic states across 25+ PB of data.

Embeddings Ingestion ➔
🧠 PHASE 03

BioHive Supercomputing

Foundation models parse structural signatures to identify untargeted disease mechanisms and infer chemical activity.

In Silico Predictions ➔
🎯 PHASE 04

Validation & Synthesis

Generative candidates loop right back into wet lab assay confirmation, continuously validating and improving model fidelity.

Closed-Loop Reset ↺
COMPUTE ENGINE TOP500 SUPERCOMPUTER

BioHive-2: Purpose-Built for Biological Modeling

Powered by state-of-the-art GPU clusters, BioHive-2 calculates relational maps of biology and chemistry. It transforms raw phenotypic imagery into multidimensional vector embeddings, exposing causal connections invisible to human researchers.

500+ H100
Accelerated GPU cluster nodes
Phenom-2
Foundational vision-biology model
BIOLOGICAL FOUNDATION HIGH-DIMENSIONAL MAPS

From Cellular Morphology to Precision Leads

Instead of testing one isolated hypothesis at a time, our phenomics engine tests every perturbation against unbiased cellular markers. We evaluate millions of chemical compounds against complex disease models simultaneously.

🔬
Unbiased Phenotypic Profiling

Fluorescence microscopy staining six cellular sub-compartments to detect subtle phenotypic shifts.

🧪
Generative Molecular Design

In silico molecular generation conditioned to reverse disease phenotypes while optimizing ADMET parameters.

📊
De-risked Target Discovery

Machine learning algorithms unmask multi-target polypharmacology and off-target liabilities early.

Operating System Status: Online See Pipeline Outcomes ➔
💊 THERAPEUTIC PIPELINE

Clinical & Preclinical Programs

Proprietary and partnered candidates accelerated through the Recursion OS, spanning oncology, rare monogenic diseases, and neuroscience.

7
Displayed
5
Phase I/II
2
IND-Ready
REC-994 Wholly-Owned
Small molecule anti-oxidant pathway
Cerebral Cavernous Malformation (CCM)
Target: Cellular Junction Stability
Complete
Complete
Complete
Phase 2 Ongoing
REC-2282 Wholly-Owned
CNS-penetrant Pan-HDAC Inhibitor
Neurofibromatosis Type 2 (NF2)
Target: Meningiomas & Schwannomas
Complete
Complete
Complete
Phase 2 Poplar
REC-4881 Wholly-Owned
Allosteric MEK1/2 Inhibitor
Familial Adenomatous Polyposis (FAP) / AXIN1 HCC
Target: APC / Beta-Catenin Suppression
Complete
Complete
Complete
Phase 2 LILAC
REC-3964 First-In-Class
Non-antibiotic Toxin B Glucosyltransferase Inhibitor
Clostridioides difficile Infection (CDI)
Target: Intestinal Epithelial Protection
Complete
Complete
Phase 1 Complete
Phase 2 Planned
RBM-007 Partnered
Novel Synthetic Lethality Compound
Target-Selected Solid Tumors
Partner: Top-Tier Biopharma Co-Dev
Complete
IND Enabled
Dosing Q3
Future
REC-1045 Wholly-Owned
Microglial Homeostasis Modulator
Neurodegenerative Tauopathies
Target: Neuroinflammatory Signaling Cascade
Complete
IND-Enabling
Planned
Planned
REC-6112 Wholly-Owned
Phenotypic Fibrotic Cascade Inhibitor
Systemic Sclerosis & Organ Fibrosis
Target: Fibroblast Transdifferentiation
Complete
In Vivo Validated
Planned
Planned
Active Milestone Stage Completed Planned Next Stage
Request Confidential Pipeline Dossier ↗
🔬 HIGH-DIMENSIONAL PHENOMICS & ML

Decoding Cellular Perturbations with Foundation Vision Models

We capture biology in its full uncompressed complexity. By treating cell morphology as a computable information layer, our neural networks isolate subtle therapeutic signals across billions of phenotypic features.

🧪
Channel Fluorescent Staining

Cell Painting at Scale

Simultaneously tagging 8 cellular sub-structures (nucleus, mitochondria, endoplasmic reticulum, actin, golgi, RNA, plasma membrane) to generate rich, unbiased morphological fingerprint matrices for every chemical or genetic intervention.

Staining Resolution 0.32 µm/pixel
🧠
Self-Supervised Vision AI

Phenom-2 Foundation Model

Trained on hundreds of millions of biological images without human labels. Phenom-2 maps raw pixels directly into low-dimensional latent embeddings, revealing cryptic disease states and predicting chemical mode-of-action in seconds.

Latent Vector Dimension 1,024 Dense Feats
⚡
In Silico Property Engine

Target-Agnostic Lead Gen

Generative diffusion models construct novel chemical scaffolds conditioned specifically to reverse diseased morphology vectors toward healthy baselines, simultaneously multi-task optimizing ADMET and selectivity.

Virtual Library Screen 100B+ Molecules
📊 IN SITU PHENOTYPIC INFERENCE

Unraveling Disease Phenotypes into Causal Vectors

Standard drug discovery measures only target binding affinity in isolation. Our platform evaluates global cellular health, identifying molecules that rescue cellular dysfunction across thousands of simultaneous morphological checkpoints.

Feature Extraction Accuracy 99.4% ROC-AUC
Target Disambiguation Speed 10x Industry Average
Off-Target Toxicity Prediction Identified in Preclinical
DATASET: MAP-HUMAN-HEK293 // V8.2 RESCUE PHENOTYPE DETECTED (Z-SCORE: 4.82)
Nuclei Granularity Normalized
Mito Organization 89% Restored
Cytoskeleton Stress -76.4% Decline
🤖 INDUSTRIALIZED AUTOMATION

Massively Parallel Robotic Wet Labs

Drug discovery can no longer rely on artisanal manual pipetting. We engineered fully autonomous bio-manufacturing cleanrooms executing millions of standardized cellular assays weekly, fueling computational learning loops with pristine high-dimensional data.

ALL AUTOMATION CLUSTERS NOMINAL FACILITY: SALT LAKE CITY CLEANROOM 04
Weekly Microplates 55,000+
Liquid Dispense Precision ± 0.5 nL
Daily Terabytes Imaged 12.8 TB/day
Human Hands In Cleanroom 0 (Autonomous)
🧪

Acoustic Liquid Ejection

Touchless soundwave fluid transfer handles fragile human biological reagents and compound solutions at nanoliter precision without tip shearing, cross-contamination, or plastic consumables waste.

Nanoliter Acoustic Precision ➔
🦾

Automated Plate Orchestration

Multi-axis robotic arms route 1,536-well and 384-well microplates continuously between computerized incubators, automated stainers, and laser confocal high-throughput automated microscopes.

24/7 Uninterrupted Operation ➔
🔄

Continuous Model Ingestion

Raw microscopy data feeds real-time into local edge servers for feature normalizations before streaming directly to BioHive supercomputers, ensuring rapid in silico validation cycles within 72 hours.

Tight Feedback Integration ➔
🏥 CLINICAL TRIALS GATEWAY

Active Studies & Patient Enrollment

Advancing precision medicines through patient-centric, biomarker-stratified clinical trials. Access protocol synopses, site maps, and screening inclusion criteria for investigators and participants.

Physician Portal ➔ Patient Assistance
Enrolling NCT05085561

The SYCAMORE Phase 2 Trial

Compound: REC-994 // Cerebral Cavernous Malformation (CCM)

A randomized, double-blind, placebo-controlled clinical trial evaluating the safety, pharmacokinetics, and efficacy of REC-994 in patients with symptomatic CCM lesion burden.

Study Design: Phase 2, 1:1:1 Dose Escalation
Primary Biomarker: Hemosiderin Volume (QSM-MRI)
Sites Enrolling: 18 Academic Centers (US/EU)
Review Protocol Synopsis ↗ Age: 18 - 65
Active Cohorts NCT05130866

The POPLAR Phase 2/3 Study

Compound: REC-2282 // Neurofibromatosis Type 2 (NF2)

Multi-center Phase 2/3 adaptive clinical trial to assess the safety and antitumoral activity of orally administered REC-2282 in participants with progressive NF2 meningiomas.

Study Design: Seamless Phase 2/3 Adaptive
Stratification: Volumetric MRI Tumor Growth
Sites Enrolling: 24 Specialist Centers
Review Protocol Synopsis ↗ Age: 12+ (Adolescent+Adult)
Enrolling NCT05552755

The LILAC Phase 2 Trial

Compound: REC-4881 // Familial Adenomatous Polyposis (FAP)

Evaluating the regression of duodenal and lower gastrointestinal polyps in adult patients with classical or attenuated FAP receiving the allosteric MEK1/2 modulator REC-4881.

Study Design: Open-Label Efficacy & Safety
Primary Endpoint: Spigelman Stage Shift / Polyp Count
Sites Enrolling: 14 High-Volume GI Centers
Review Protocol Synopsis ↗ Age: 18 - 70

Are you a clinician, researcher, or patient?

Our specialized clinical team provides secure transfer of study documentation, pre-screening inclusion guidance, and dedicated site coordinators for participating hospitals.

🤝 ENTERPRISE ALLIANCES

Biopharma Partnerships at Industrial Scale

We partner with leading global pharmaceutical innovators to solve intractable target biology, optimize lead generation, and accelerate transformative therapies from computational hits to clinic.

Neuroscience & Oncology Up to $12B+

Roche & Genentech

Multi-Year Strategic Collaboration

Leveraging the Recursion OS to decode novel therapeutic targets across 40 programs spanning neuroscience and oncology indications, unlocking phenotypes in previously inaccessible disease spaces.

Active Programs: 10+ Tier 1 Multi-Target
Fibrotic Diseases Up to $1B+

Bayer AG

Target Discovery & Small Molecules

Fusing human cellular disease models with automated high-throughput phenotypic screening to systematically uncover and de-risk novel targets driving pulmonary and cardiac fibrotic mechanisms.

Focus: Organ Fibrosis Preclinical Success
Supercomputing & AI BioHive-2 Infra

NVIDIA

Generative Biology AI Co-Development

Direct collaboration accelerating foundational biological models, optimizing Phenom-2 on NVIDIA DGX SuperPOD systems, and deploying generative models through NVIDIA BioNeMo microservices.

Compute: 500+ H100 GPUs Generative Biology
⚙️ COLLABORATION FRAMEWORKS

Flexible Co-Development & Out-Licensing Models

Whether accessing our proprietary Phase 2 clinical assets or initiating new multi-target discovery pipelines using our robotic facilities and foundational AI models, we tailor structures to partner strategic objectives.

Initiate Partnership Dialogue ↗ View Available Out-Licenses
🎯

Target Identification

Interrogate intractable disease pathways using whole-genome CRISPR and massive high-content imaging datasets.

💊

Clinical Asset Licensing

Acquire global co-development and commercialization rights for active clinical oncology and rare-disease molecules.

🧪

Generative Chemistry

Design bespoke chemical scaffolds to navigate tricky target pharmacology with high selectivity and drug-like properties.

⚡

Phenomic Screening APIs

Screen external proprietary libraries through Recursion automated assays with algorithmic target deconvolution.

📈 SHAREHOLDER PORTAL

Investor Relations

Real-time financial telemetry, SEC regulatory disclosures, upcoming conference webcasts, and corporate governance records for shareholders and institutions.

NASDAQ: RXRX $9.42
+4.18% ↗ Market Cap: ~$2.4B

📅 Events & Presentations

Export Calendar
WEBCAST NOV 14, 2024 · 4:30 PM ET

Q3 2024 Financial Results & Business Update

Conference call and live audio webcast with Q&A session

CONFERENCE DEC 03, 2024 · 10:15 AM PT

Piper Sandler 36th Annual Healthcare Conference

Fireside chat with CEO Chris Gibson, Ph.D.

SYMPOSIUM JAN 13, 2025 · 8:00 AM PT

43rd Annual J.P. Morgan Healthcare Conference

Corporate presentation and pipeline milestone readouts

Sign up for automatic investor calendar alerts Subscribe

📄 Recent SEC Filings

View All ↗
Form 10-Q (Quarterly Report) Quarter ended September 30, 2024
PDF (2.4 MB)
Form 8-K (Current Report) Filed October 18, 2024 · Corporate Update
HTML
Form 8-K (Milestone Achievement) Filed September 04, 2024 · Roche Milestone
PDF (840 KB)
Form 10-K (Annual Report) Fiscal year ended December 31, 2023
PDF (6.8 MB)
Investor Relations Contact: ir@recursion.com
⚖️ Corporate Governance Charters, Bylaws & Ethics
👥 Board of Directors Independent Oversight
🌱 ESG Impact Report Sustainability & Responsibility
📬 Analyst Coverage 14 Institutional Brokerages
📚 PEER-REVIEWED RIGOR & OPEN SCIENCE

Publications & Open Datasets

Empirical validation of our digital biology platform across leading scientific journals, machine learning conferences, and publicly released benchmark datasets for the computational biology community.

RxRx.ai Portal ↗
NATURE BIOTECHNOLOGY PUBLISHED MARCH 2024 · PEER-REVIEWED

High-Content Phenotypic Profiling at Multi-Million Well Scale Uncovers Novel Disease Modulators

Demonstrating how unbiased computer vision models applied to 5-channel cell painting across 2.2 million genetic and chemical perturbations uncover previously unknown functional connections between disparate signaling pathways.

DOI: 10.1038/s41587-024-00000-x Citations: 140+ Download Open Access PDF ➔
Key Paper Takeaways
  • ✓ 10x improvement in rescue hit-rate over standard target binding assays.
  • ✓ Reconstructed complete metabolic cascades without manual pathway curation.
  • ✓ Validated efficacy in animal models for two novel neuro-oncology targets.
NeurIPS 2023 · Spotlight Machine Learning

Phenom-1: Self-Supervised Vision Transformers for Cellular Phenotyping at Supercomputing Scale

Introduces a vision transformer architecture pre-trained on 400M microscopy images, achieving state-of-the-art representation learning on cell morphology and genetic knockout classification.

Gibson et al. View Paper (arXiv) ↗
Nature Communications Translational Medicine

Identification of Allosteric MEK Inhibitors via High-Throughput Structural Phenomics in FAP Models

Preclinical discovery and validation framework behind REC-4881, revealing mucosal restoration mechanisms in APC-deficient adenomatous polyposis organoid models.

Wong, Adams et al. Read Full Text ↗
ICML 2024 · Workshop Computational Biology

Batch Effect Normalization in Petabyte-Scale Fluorescent Cellular Datasets

Novel spherical contrastive learning algorithms to mathematically disentangle non-biological experimental plate noise from genuine phenotypic drug effects.

Sypetkowski et al. View PDF ↗
Cell Systems Systems Biology

Comprehensive Mapping of Small-Molecule Polypharmacology via High-Content Cell Morphology

Demonstrating systematic prediction of secondary target engagement and unpredicted in vivo toxicities from early high-throughput in vitro cellular profiles.

Kemp, et al. Read Full Text ↗
🌐 Community Open Data Initiative

The RxRx Open Biological Datasets (RxRx1, RxRx2, RxRx3)

Recursion has open-sourced hundreds of gigabytes of fluorescent cell imaging datasets to advance academic ML research. Used by universities and AI researchers globally for competitive benchmarks.

Access Datasets on RxRx.ai ↗ GitHub Code Repository