Decoding Biology To Radically Improve Lives End of Q3 2023
2 Disclaimers This presentation and any accompanying discussion and documents contain information that includes or is based upon "forward-looking statements" within the meaning of the Securities Litigation Reform Act of 1995. These forward-looking statements are based on our current expectations, estimates and projections about our industry and our company, management's beliefs and certain assumptions we have made. The words “plan,” “anticipate,” “believe,” “continue,” “estimate,” “expect,” “intend,” “may,” “will” and similar expressions are intended to identify forward-looking statements. Forward-looking statements made in this presentation include outcomes and benefits expected from the Tempus partnership, including our ability to leverage the datasets acquired through the license agreement into increased machine learning capabilities and accelerate clinical trial enrollment; our planned expansion of the BioHive supercomputer capabilities; outcomes and benefits from licenses, partnerships and collaborations, including option exercises by partners and additional partnerships and ability to house tools on the BioNeMo Marketplace; the occurrence or realization of any near- or medium-term potential milestones; the initiation, timing, progress, results, and cost of our research and development programs and our current and future preclinical and clinical studies, including timelines for data readouts, the potential size of the market opportunity for our drug candidates; our ability to identify viable new drug candidates for clinical development and the accelerating rate at which we expect to identify such candidates; our expectation that the assets that will drive the most value for us are those that we will identify in the future using our datasets and tools, and many others. Forward-looking statements made in this presentation are neither historical facts nor assurances of future performance, are subject to significant risks and uncertainties, and may not occur as actual results could differ materially and adversely from those anticipated or implied in the forward-looking statements. For a discussion of factors that could affect our business, please refer to the "Risk Factors" sections in our filings with the U.S. Securities and Exchange Commission, including our Annual Report for the Fiscal Year ended December 31, 2022, on Form 10-K and our most recent Quarterly Report on Form 10-Q. This presentation does not purport to contain all the information that may be required to make a full analysis of the subject matter. We undertake no obligation to correct or update any forward-looking statements, whether as a result of new information, future events or otherwise. Certain information contained in this presentation relates to or is based on studies, publications, surveys and other data obtained from third-party sources and the company’s own internal estimates and research. While the company believes these third-party sources to be reliable as of the date of this presentation, it has not independently verified, and makes no representation as to the adequacy, fairness, accuracy or completeness of, any information obtained from third-party sources. In addition, all of the market data included in this presentation involves a number of assumptions and limitations, and there can be no guarantee as to the accuracy or reliability of such assumptions. Finally, while the company believes its own internal research is reliable, such research has not been verified by any independent source. Any non-Recursion logos or trademarks included herein are the property of the owners thereof and are used for reference purposes only.
3 Real (atoms) Digital (bits) 1 Profile Systems Capture high-dimensional data to create a digital record of reality (things, places, preferences, etc.) Data Aggregate and organize data to create digital maps of reality Algorithms Navigate digital maps to predict novel relationships, then try them in reality 3 2 1 2 3 Since our founding, we have been building virtuous cycles of atoms & bits to accelerate & improve drug discovery
4 Trademarks are the property of their respective owners and used for informational purposes only. Reproducibility Crisis Multiple studies have shown that the vast majority of published academic literature cannot be recapitulated Analog Standard The fax machine is alive and well in medicine, while in biopharma, study results from CROs are still often reported as PDFs or scanned printouts Siloed Data in Pharma The culture of biopharma has led to 100s of petabytes of scientific data being stored on a project-by-project basis without the meta-data or annotation needed to relate it to other projects or questions in biology ! ! ! ! ! ! ! ! ! ! Our approach breaks down the data roadblocks that challenge the traditional Biopharma industry
Improved and scaled clinical pipeline Recursion OS Profile Systems: We have built and continue to scale among the world’s most prolific automated wet labs 5 Data: Each week we digitize millions of our own experiments across multiple layers of biology from cell to animal ALGORITHMS: We own and operate one of the fastest supercomputers on earth, allowing us to train LLMs & FMs fit for the purpose of drug discovery To truly unlock the enormous promise of TechBio, an integrated approach combining wet-lab, the right data & powerful computation is imperative
Updates reinforce Recursion’s position as a leader at the intersection of scaled biology and compute • 5-year partnership providing access to multi-modal data for >100k oncology patients • Ability to build causal AI models to tune Recursion OS • More precisely aim clinical programs towards the right patients • Expanding Biohive-1 supercomputer with NVIDIA to quadruple in-house compute power • Enable continued acceleration of our efforts to train foundation models and LLMs for purposes of drug discovery • Partnership evolving to focus on precision oncology in support of Bayer’s updated corporate strategy • Per program milestones up >1.5X from initial collaboration A leap forward in our vision 6 Data Computation Pipeline
Update 1: Tempus Partnership
+ • $160M paid by Recursion to Tempus in cash or equity, at our election, in increasing annual increments over five years, beginning with $22M of equity to be issued later this year • Expected to accelerate model deployment, linking molecular data with outcomes • Expected to enhance program translation as well as identification and enrollment of patients with higher probability of clinical response • Provides preferential access to DNA / RNA sequencing datasets tied to clinical records for >100,000 patients for the purpose of training causal AI models for therapeutic development Proposed partnership accelerates clinical platform capabilities with ~50 PB of proprietary biology, chemistry, and translational precision medicine data purpose-built for AI / ML Recursion to partner with Tempus 8
Creating another potential virtuous cycle: Recursion will direct biological exploration by the Recursion OS based on scaled patient data as well as drive programs into the clinic with biomarker and patient stratification insights at an unparalleled pace 20PB+ of data immediately available to Recursion CLINICAL TBA DNAseq 150K RNAseq 0K 10K DNA and RNA CLINICAL 6M+ DNAseq 600K+ RNAseq 160K+ Partnership provides preferential access to one of the world’s largest and fastest growing libraries of real-world multimodal oncology data 9TCGA AACR-GENIE
10 Partnership will create among the most comprehensive set of biological data layers in the industry Image adapted from D’Orazio, M., et al. Nature Scientific Reports 2022 RNA Level (Cell and Population-Scale) Transcriptomics DNA Level (Population-Scale) Genomics Protein Level Proteomics Organism Level InVivomics Cell Level Phenotypes Phenomics Built, scaled or partnered AspirationalExploratory Like digital maps of Earth, connections within and between layers add useful context. Similarly, Recursion is mapping multi-omic layers of biology and identifying connections within and between layers to decode biology at scale. Clinical Biomarker & Health Records Real-World Patient Data
11 Recursion Data Universe: >25 PB of proprietary biological and chemical data, spanning phenomics, transcriptomics, inVivomics, and more • We believe this is one of the largest such datasets fit for the purpose of training large-scale ML models in biology • RXRX3: CRISPR knockouts of most of the human genome, 1,600 FDA approved / commercially available bioactive compounds • We believe this is the largest public dataset of its kind, <1% of Recursion Data Universe and what Recursion can generate in ~1 week Phenomics data spanning >200 million experiments and ~50 human cell types InVivomics Video Data >500K Transcriptomics Experiments ADMET & More R e cu rsio n D ata U n ive rse , >2 5 P B Preferred access to de-identified, multi-modal clinical and molecular records through multi-year licensing agreement A cce ssib le P ro p rietary D ata, ~5 0 P B Publicly- released RXRX3 <1% Preferential access to >20 PB of real-world patient data • Includes access to more than 100,000 oncology patient’s de-identified records, DNA sequencing, RNA sequencing, and clinical outcomes on which we can train causal AI models to: • Tune our Recursion OS for increased translatability • Better target clinical programs to the right patients New capabilities accelerate scale & enhance Recursion’s relatable data differentiation
12 Collaboration provides preferential access to multi-modal data to enhance precision medicine, translation and trial design TranslationCompound OptimizationHit & Target ValidationPatient Connectivity & Novelty Chemical property predictions Compound selection with ML DMPK experiments Phenomic efficacy InVivomic Prioritization & Digital Tolerability InVivomic Efficacy Preclinical assessment ~2M physical compounds & ~36B compounds with target predictions from Matchmaker ML system ~50 human cell types including whole genome CRISPR-based knockouts enable reverse- genetic causal ML models LLM-based automated IND generation LLM-based automated patent writing Preferential access to Tempus data (>20 PBs) enables training of patient-centric ML models LLMs query global knowledge for program ideation & evaluation Combined models and tools will enable automated generation of novel, patient-specific hypotheses Preferential access to Tempus data enables both single and complex biomarker identification strategy for patient stratification IND Enabling Studies Clinical Development Predicted target binding Scaled Phenomics Scaled Transcriptomics Chemical tractability
Update 2: Supercomputer Expansion
Computational Tools Expand BioHive-1 from: • 320 A100s… …to include an additional • 504 H100s With operations beginning H1 2024 Likely to be the highest performing compute cluster owned and operated by any biopharma company on earth and among the top 50 compute clusters on the Top500 list 14
The combination of scaled data generation and accelerated computing is a key to advancing biological ML World’s largest phenomic foundation model for biology (PHENOM-1) 15 Increasingly powerful foundation models for chemistry Orchestration with custom LLMs to minimize toil and bias from discovery at scale Vision: Autonomous agents automatically explore, predict and execute our workflows to discover and develop medicines with hand-off to humans at later stages Advancements driven by increasingly scaled data generation and compute; leading to reductions in human bias at every step 15
Update 3: Bayer Partnership Transformation
17 Update of existing collaboration to exploit Recursion OS advancements and align with Bayer’s strategic interest in precision oncology Go-forward collaboration • Re-aligned focus to deliver on Bayer strategic objectives in oncology • Up to 7 new Projects anticipated • >1.5x increase in per program economics • Designed to leverage advancements in Recursion OS platform since partnership inception • Foundation models and LLMs deployed to identify novel targets of interest • Industrialized workflows prosecute programs with increasing likelihood of translation and minimal human bias • Application of digital chemistry tooling from acquisitions of Cyclica and Valence Original Collaboration (announced Sept. 2020) Focus: Fibrosis
In Brief: The Recursion Value Proposition
Recursion leading a new TechBio sector at the intersection of technology and biology 19Adapted from Scannell, J et al (2012). Diagnosing the decline in pharmaceutical R&D efficiency. Nat Rev Drug Discov, 11, 191-200. 0.1 1 10 50B 100M 500K 1K 20201962 1970 1980 1990 2000 2010 Compute Power (Transistors per Microprocessor)NMEs per $B spent (inflation adjusted) Moore’s Law: Computing power becomes faster and less expensive over time Eroom’s Law: Drug discovery is becoming slower and more expensive over time Opportunity
20 The Recursion OS today: Industrializing drug discovery to transform BioTech into TechBio TranslationCompound OptimizationHit & Target ValidationPatient Connectivity & Novelty Chemical property predictions Compound selection with ML DMPK experiments Phenomic efficacy InVivomic Prioritization & Digital Tolerability InVivomic Efficacy Preclinical assessment ~2M physical compounds & ~36B compounds with target predictions from Matchmaker ML system ~50 human cell types including whole genome CRISPR-based knockouts enable reverse- genetic causal ML models Predicted target binding Scaled Phenomics Scaled Transcriptomics Chemical tractability LLM-based automated IND generation LLM-based automated patent writing Preferential access to Tempus data (>20 PBs) enables training of patient-centric ML models LLMs query global knowledge for program ideation & evaluation Combined models and tools will enable automated generation of novel, patient-specific hypotheses Preferential access to Tempus data enables both single and complex biomarker identification strategy for patient stratification IND Enabling Studies Clinical Development
Data shown is the average of all our programs since late 2017 through 2022. All industry data adapted from Paul, et al. Nature Reviews Drug Discovery. (2010) 9, 203–214 Mapping and navigating the complex systems of biology and chemistry has demonstrated leading indicators of efficiency 21 Screen — Hit ID — Validated Lead — Advanced Candidate — Development Candidate — Industry Recursion Failing faster and earlier to › › spend less › C o st t o IN D ( $ M ) 25 — 20 — 15 — 10 — 5 — 0 — Industry 40 — 30 — 20 — 10 — 0 — › and go faster 100 80 60 51 80% 75% 85% 100 64 25 64% 39% 62% 12 50% St ag e Ti m e t o V al id at ed L e ad ( m o ) Recursion Industry Recursion 8 Industry Recursion
Harnessing value with a multi-pronged capital-efficient business strategy 22 Pipeline Recursion OS Build internal pipeline in indications with potential for accelerated path to approval Pipeline Strategy Precision Oncology Rare Disease Partnerships Data Partnership Strategy Partner in complex therapeutic areas requiring large financial commitment or competitive arbitrage Leverage partner knowledge and clinical development capabilities License subsets of data and key tools Direct generation of new data internally to maximize pipeline and partnership value-drivers Data Strategy Undruggable Oncology Other large, intractable areas of biology (e.g. CV/Met) Neuroscience* Licensing Augment Recursion OS *Includes a single oncology indication from our Roche and Genentech collaboration.
PreclinicalLate Discovery Oncology Rare & Other Therapeutic Area Indication AXIN1 or APC MUTANT CANCERS (AXIN1 or APC mutant cancers; est. 65K) MYC-DRIVEN ONCOLOGY (MYC; est. 54K5) CLOSTRIDIOIDES DIFFICILE INFECTION (est. 730K) CEREBRAL CAVERNOUS MALFORMATION (CCM; est. 360K1) NEUROFIBROMATOSIS TYPE 2 (NF2; est. 33K2) FAMILIAL ADENOMATOUS POLYPOSIS (APC; est. 50K3) CANCER IMMUNOTHERAPY, TARGET ALPHA (Multiple; est. 72K4) Phase 1 Phase 2 Phase 3 23 More than a dozen additional early discovery and research programs in oncology or with our partners – first program already optioned by Roche-Genentech in GI-oncology Our pipeline reflects the scale and breadth of our approach CANCER IMMUNOTHERAPY, TARGET DELTA (Multiple; est. 88K4) HR-PROFICIENT OVARIAN CANCER, RBM39 (HR-proficient ovarian cancer; est. 13K) All populations defined above are US and EU5 incidence unless otherwise noted. EU5 is defined as France, Germany, Italy, Spain and UK. (1) Prevalence for hereditary and sporadic symptomatic population. (2) Annual US and EU5 incidence for all NF2-driven meningiomas. (3) Prevalence for adult and pediatric population. (4) Our program has the potential to address several indications in this space. (5) Our program has the potential to address several indications driven by MYC alterations, totaling 54,000 patients in the US and EU5 annually. We have not finalized a target product profile for a specific indication.
Our existing and recent partnerships represent some of the most significant scientific collaborations in TechBio across biopharma and tech Trademarks are the property of their respective owners and used for informational purposes only. 24 Undruggable oncology targets • $30M upfront and $50M equity investment • Increased per program milestones which may be up to $1.5B for up to 7 oncology programs • Mid single-digit royalties on net sales • Recursion owns all algorithmic improvements Neuroscience and a single oncology indication • $150M upfront and up to or exceeding $500M in research milestones and data usage options • Up to or exceeding $300M in possible milestones per program for up to 40 programs • First program already optioned • Mid to high single-digit tiered royalties on net sales Computation and ML/AI • $50M equity investment • Partnership on advanced computation (e.g., foundation model development) • Priority access to compute hardware or DGXCloud Resources • Potential to house Recursion Tools on NVIDIA’s BioNeMo Marketplace Real-world data access • Preferential access to >20 PBs of Tempus real-world, multi- modal oncology data, including DNA/RNA sequencing and clinical outcome data for more than 100,000 patients • Ability to train causal AI models with utility in target discovery, biomarker development & patient selection • Opportunity to accelerate clinical trial enrolment through potential access to broad clinical network Therapeutic discovery collaborations Technology and data access collaborations Announced Dec. 2021 Announced Nov. 2023 Announced July 2023 Announced Sept. 2020 Announced Nov. 2023
How we build maps of biology and chemistry to turn drug discovery into a search problem
26 The Recursion OS today: Industrializing drug discovery to transform BioTech into TechBio TranslationCompound OptimizationHit & Target ValidationPatient Connectivity & Novelty Chemical property predictions Compound selection with ML DMPK experiments Phenomic efficacy InVivomic Prioritization & Digital Tolerability InVivomic Efficacy Preclinical assessment ~2M physical compounds & ~36B compounds with target predictions from Matchmaker ML system ~50 human cell types including whole genome CRISPR-based knockouts enable reverse- genetic causal ML models Predicted target binding Scaled Phenomics Scaled Transcriptomics Chemical tractability LLM-based automated IND generation LLM-based automated patent writing Preferential access to Tempus data (>20 PBs) enables training of patient-centric ML models LLMs query global knowledge for program ideation & evaluation Combined models and tools will enable automated generation of novel, patient-specific hypotheses Preferential access to Tempus data enables both single and complex biomarker identification strategy for patient stratification IND Enabling Studies Clinical Development
27 27 TranslationCompound OptimizationHit & Target ValidationPatient Connectivity & Novelty State-of-the-art LLMs query global knowledge for program ideation & evaluation Rapidly scaling to 1000s of new differentiated program ideas Large language models (LLMs) evaluate complex opportunities at scale Differentiation & Impact Novel map insights and rapid disease research e.g., Uncover which of our 300M+ gene-gene relationships are unique to our Maps Automation & Scale High-throughput LLMs reduce manual research load & human bias e.g., Our 250,000 tokens/min LLM capacity BioHive-1 is a global TOP500 supercomputer Model Accuracy Correct Incorrect GPT-NeoX 47% 53% Dolly2.0 49% 51% GPT4.0 on Azure 80% 20% LLaMA 2 Under evaluation Ability to bridge to ~2M physical compounds & ~36B compounds with target prediction Combined models and tools will enable automated generation of novel, patient-specific hypotheses Our LLMs quickly distill the most promising novel ideas from >5 trillion relationship search space Preferential access to Tempus data (>20 PBs) enables training of patient- centric ML models ~50 human cell types including whole genome CRISPR-based knockouts enable reverse-genetic causal ML models
28 28 TranslationCompound OptimizationHit & Target ValidationPatient Connectivity & Novelty Rapidly scaling to 1000s of new differentiated program ideas Hundreds of compound hits per gene from our Maps >100 programs generated in H1 2023 Automated data package Distilled insights for rapid decision making Predicted target binding Phenomics Transcriptomics Chemical tractability Automatic validation of map insights: we rapidly confirm novel predictions from our maps with automated, standardized, scaled -omics
29 29 TranslationCompound OptimizationHit & Target ValidationPatient Connectivity & Novelty Chemical property predictions ML predicted protein-ligand interactions for ~36 billion compounds and ML predictions for new compounds that become part of our library Compound selection with ML Semi-generative and multi- objective generative chemistry models drive compound design DMPK experiments A highly automated DMPK module executes 3 critical assays across human and rat contexts Phenomic efficacy Compound potency and selectivity are rapidly measured using scaled phenomics Automated loops of in silico predictions & robotic experiments Automated data package Loops of experimental data & ML predictions rapidly accelerate hit to lead and lead optimization
30 30 TranslationCompound OptimizationHit & Target ValidationPatient Connectivity & Novelty InVivomic Prioritization & Digital Tolerability Rat and mouse studies with ML-based selection of optimal compound and dose from videos Preclinical assessment Best options from Compound Optimization Poorly tolerated compounds kicked back for further Compound Optimization Automated data package Not tolerated Tolerated ML evaluation of mice against >10 liabilities. InVivomic Efficacy InVivomic ML non-invasively enables phenotype induction confirmation and efficacy determination Automated data package InVivomics improves whole organism understanding to rapidly translate programs towards the clinic
31 Recursion OS LLM-based orchestration and autonomous exploration LLMs query global knowledge for program ideation & evaluation Combined models and tools will enable automated generation of novel, patient-specific hypotheses Custom AI Chemist LLM minimizes human toil and bias approaching later-stage development Data packages automatically generated to inform stage-gate decision making LLM-based automated IND generation LLM-based automated patent writing Active-learning informed chemistry navigate billions of synthesizable molecules across chemical space to direct more efficient DMTA cycles Roadmap: Integration and orchestration of tools with LLMs/API Calls to create super-empowered scientists & facilitate autonomous exploration TranslationCompound OptimizationHit & Target ValidationPatient Connectivity & Novelty Chemical property predictions Compound selection with ML DMPK experiments Phenomic efficacy InVivomic Prioritization & Digital Tolerability Preclinical assessment ~2M physical compounds & ~36B compounds with target predictions from Matchmaker ML system ~50 human cell types including whole genome CRISPR-based knockouts enable reverse- genetic causal ML models Predicted target binding Scaled Phenomics Scaled Transcriptomics Chemical tractability IND generation Patent writing Preferential access to Tempus data (>20 PBs) enables training of patient-centric ML models IND Enabling Studies Clinical Development InVivomic Efficacy Preferential access to Tempus data enables both single and complex biomarker identification strategy for patient stratification
32 Our virtuous cycles of atoms and bits are already leading to first-in-disease development and beyond
Julia – living with CCM 33 SYCAMORE Clinical Trial : REC-994 for CCM Phase 2 Fully Enrolled Clinical: CCM Symptomatic US + EU5, >1 million patients worldwide live with these lesions today PREVALENCE & STANDARD OF CARE CAUSE LOF mutations in genes CCM1, CCM2 & CCM3, key for maintaining the structural integrity of the vasculature due to unknown mechanisms PATHOPHYSIOLOGY & REASON TO BELIEVE Efficacy in Recursion OS as well as functional validation via scavenging of massive superoxide accumulation in cellular models; reduction in lesion number with chronic administration in mice KEY ELEMENTS • Targeting sporadic and familial symptomatic CCM patients with CCM1, CCM2, and CCM3 mutations • Superoxide scavenger, small molecule • Phase 2 trial initiated in Q1 2022 • US & EU Orphan Drug Designation • Oral dosing ~360,000 Vascular malformations of the CNS leading to focal neurological deficits, hemorrhage and other symptomsNo approved therapy • Most patients receive no treatment or only symptomatic therapy • Surgical resection or stereotactic radiosurgery not always feasible because of location and is not curative >5x larger US patient population than other rare diseases like Cystic Fibrosis (>31k patients) Vascular malformations (cavernomas)
34 SYCAMORE Clinical Trial : REC-994 for CCM Phase 2 Fully Enrolled Phase 2 trial initiated in Q1 2022 Source: https://www.clinicaltrials.gov/ct2/show/NCT05130866?term=recursion&draw=2&rank=3; https://www.SycamoreCCM.com/ Screening & Randomization 1:1:1 Treatment Follow-up Outcome Measures Enrollment Criteria • MRI-confirmed CCM lesion(s) • Familial or sporadic • Symptoms directly related to CCM • Primary: Safety and tolerability • Adverse events & symptoms • Secondary: Efficacy • Clinician-measured outcomes (CGI and PGI) • Imaging of CCM lesions – number, size & rate of change • Impact of acute stroke (mRS, NIHSS) • Patient reported outcomes (SMSS, PROMIS-29, CCM HI, symptom questionnaires) • Exploratory: Biomarkers • Enrollment is complete • Vast majority of participants have completed 12 months of treatment and entered long-term extension study • Top-line data expected H2 2024 Trial Update Clinical: CCM 34 400mg 200mg Placebo Visits: Days 1 & 2 Months 1, 3, 6, 9 & 12 Enroll ~60 Extension Study12 Months Treatment Period
35 POPLAR Clinical Trial : REC-2282 for NF2 Part A Underway Clinical: NF2 LOF mutations in NF2 tumor suppressor gene, leading to deficiencies in the tumor suppressor protein merlin PATHOPHYSIOLOGY & REASON TO BELIEVE Efficacy in Recursion OS, cellular, and animal models; suppression of aberrant ERK, AKT, and S6 pathway activation in a Phase 1 PD Study in NF2 patient tumors KEY ELEMENTS • Targeting familial and sporadic NF2 meningioma patients • HDAC inhibitor, small molecule • Oral dosing • Phase 2/3 trial initiated in Q2 2022 • Fast-Track and US & EU Orphan Drug Designation Inherited rare CNS tumor syndrome leading to loss of hearing and mobility, other focal neurologic deficits No approved therapy • There are no approved drugs for NF2 • Surgery is standard of care (when feasible) • Location may make complete resection untenable, leading to hearing loss, facial paralysis, poor balance and visual difficulty Treatable US + EU~33,000 PREVALENCE & STANDARD OF CARE CAUSE Ricki – living with NF2 Intracranial meningiomas
POPLAR Clinical Trial : REC-2282 for NF2 Part A Underway Phase 2/3 trial initiated in Q2 2022 Outcome Measures Enrollment Criteria • MRI-confirmed progressive meningioma • Either of the below • Sporadic meningioma with confirmed NF2 mutation • Confirmed diagnosis of NF2 disease • Primary: Safety and tolerability • Progression-free survival • Time to progression • Duration of response • Overall response rate https://clinicaltrials.gov/ct2/show/NCT05130866 Clinical: NF2 Screening & Randomization 1:1 Treatment Follow-up Agreement on Phase 3 registration plans FDA Mtg Phase 2 (Cohort A) Phase 3 (Cohort B) Interim Analysis ▪ At 50% of events ▪ For Sample size re-estimation (i.e., adaptive design) • Enrollment is progressing • Safety, tolerability, PK, & preliminary efficacy expected in H2 2024 Trial Update 36 Enroll 60 26 Month Tx Period Extension Study 6-month Tx Period (Interim Analysis) Extension Study 60 mg TIW 40 mg TIW Enroll ~ 20 ▪ Go/No-go to Ph3 ▪ Safety/Tolerability ▪ PK ▪ PFS Cohort A Final Data
37 TUPELO Clinical Trial : REC-4881 for FAP Phase 2 Underway Clinical: FAP Polyps Found in Colon and Upper GI Tract Inactivating mutations in the tumor suppressor gene APC PATHOPHYSIOLOGY & REASON TO BELIEVE KEY ELEMENTS • Targeting classical FAP patients (with APC mutation) • MEK inhibitor, small molecule • Oral dosing • Phase 2 trial initiated in Q3 2022 • Fast-Track and US & EU Orphan Drug Designation Polyps throughout the GI tract with extremely high risk of malignant transformation Efficacy in the Recursion OS showed specific MEK 1/2 inhibitors had an effect in context of APC LOF. Subsequent APCmin mouse model showed potent reduction in polyps and dysplastic adenomas No approved therapy • Colectomy during adolescence (with or without removal of rectum) is standard of care • Post-colectomy, patients still at significant risk of polyps progressing to GI cancer • Significant decrease in quality-of-life post-colectomy (continued endoscopies, surgical intervention) Diagnosed US + EU5~50,000 PREVALENCE & STANDARD OF CARE CAUSE
38 TUPELO Clinical Trial : REC-4881 for FAP Phase 2 Underway Clinical: FAP Phase 2 trial initiated in Q3 2022 Outcome Measures Enrollment Criteria • Confirmed APC mutation • Post-colectomy/proctocolectomy • No GI cancer present • Polyps in either duodenum (including ampulla of vater) or rectum/pouch • Primary: • Part 1: PK • Part 2: polyp burden (% change from baseline) • Secondary: • Part 1: Safety & tolerability • Part 2: PK; PD; change from baseline in polyp number, histological grade, disease score https://clinicaltrials.gov/ct2/show/NCT05552755, protocol amendments made to enhance quality and accelerate the pace of the trial 38 Part 1 Part 2 Dose Expansion (N~30) at RP2D • Futility Assessment • Go/No-Go Single agent REC-4881 Dose Escalation • Safety • Tolerability • PK/PD Screening & Treatment Trial Update • Enrollment is progressing • Safety, tolerability, PK, & preliminary efficacy expected in H1 2025 4 mg QD (n ≤ 6) 8 mg QD (n ≤ 6) 12 mg QD (n ≤ 6) Recommended Phase 2 Dose
39 LOF mutations in AXIN1 or APC tumor suppressor genes PATHOPHYSIOLOGY & REASON TO BELIEVE Alterations in the WNT pathway are found in a wide variety of tumors and confer poor prognosis and resistance to standard of care Substantial need for developing therapeutics for patients harboring mutations in AXIN1 or APC, as these mutations are considered undruggable Treatable US + EU5~65,000 PREVALENCE & STANDARD OF CARE CAUSE Efficacy in the Recursion OS and favorable results in PDX models harboring AXIN1 or APC mutations vs wild-type leading to a significant PFS benefit in HCC and Ovarian tumors Gross morphology of HCC KEY ELEMENTS • Targeting AXIN1 or APC mutant cancers • MEK inhibitor, small molecule • Oral dosing • IND accepted by FDA • Expect to initiate Phase 2 study in late Q4 2023 or early Q1 2024 To our knowledge, REC-4881 is the only industry sponsored small molecule therapeutic designed to enroll solid tumor patients harboring mutations in AXIN1 or APC LILAC Clinical Trial : REC-4881 for AXIN1 or APC mutant cancers Clinical: AXIN1 or APC
Expect Phase 2 initiation in late Q4 2023 or early Q1 2024 Outcome Measures Enrollment Criteria • Unresectable, locally advanced, or metastatic cancers • AXIN1 or APC mutation confirmed by NGS (tissue or blood) • CRC patients must be RAS / RAF wildtype • No MEK inhibitor treatment within 2 months of initial dose • ≥ 1 prior line of therapy • ECOG PS 0-1 • Primary • Safety/tolerability • ORR (RECIST 1.1) • Secondary • PK • Additional efficacy parameters • First clinical trial for an oncology indication at Recursion • IND accepted by FDA Trial Update 40 Safety Assessment 4 mg, 8 mg, 12 mg REC-4881 QD https://clinicaltrials.gov/ct2/show/NCT06005974, protocol amendments made to enhance quality and accelerate the pace of the trial R P 2 D AXIN1 (n=10) APC (n=10) Futility Assessment AXIN1 (n=10) APC (n=10) Once 10 pts enrolled in each cohort with ≥ 1 scan post-baseline Futility Assessment Screening & Treatment Part 1 Part 2 LILAC Clinical Trial : REC-4881 PoC for AXIN1 or APC mutant cancers Clinical: AXIN1 or APC
41 Colleen – lived with rCDI C. difficile toxins from colonizing bacterium causes degradation of colon cell junction, toxin transit to bloodstream, and morbidity to host PATHOPHYSIOLOGY & REASON TO BELIEVE Highly recurrent infectious disease with severe diarrhea, colitis, and risk of toxic megacolon, sepsis, and death Diagnosed US + EU5~730,000 PREVALENCE & STANDARD OF CARE CAUSE Recursion OS identified a new chemical entity for recurrent C. difficile infection and potentially prophylaxis via glycosyl transferase inhibition with potential to be orally active KEY ELEMENTS • Phase 1 PK study complete • REC-3964 was well tolerated and all AEs were Grade 1 • Expect to initiate Phase 2 proof-of-concept study in 2024 • Selective C. diff toxin inhibitor, small molecule • Non-antibiotic approach with potential for combination with SOC and other therapies • Designed for selective antitoxin pharmacology to target infection • Phase 1 HV study complete TRIAL UPDATE Standard of care includes antibiotic therapies which can further impair gut flora, and lead to relapse Clinical Trial : REC-3964 for C. Difficile Phase 1 Study Complete Clinical: C. Difficile
42 Clinical Trial : REC-3964 for C. Difficile Phase 1 Study Complete Trial Design • Randomized, Double-blind Trial Population • Healthy Participants • SAD (n = 48) • 36 participants treated with REC-3964 • 12 participants treated with placebo • MAD (n = 42) • 34 participants treated with REC-3964 • 8 participants treated with placebo Primary Objectives ✓ Assess the safety & tolerability of SAD and MAD of REC-3964 ✓ Evaluate the PK profile of REC-3964 after single and multiple doses Clinical: C. Difficile 42 Phase 1 Topline • REC-3964 oral administration was well tolerated by all subjects tested ✓ 3% (n=1) of participants in SAD with drug-related AEs ✓ 12% (n=4) of participants in MAD with drug-related AEs ✓ All AEs were deemed Grade 1 ✓ No SAEs were observed ✓ No discontinuations related to treatment • REC-3964 exhibited a favorable PK profile ✓ Exposures (AUC) increased approximately dose-proportionally across the dose ranges tested (50 mg – 1200 mg) ✓ Half-life ranged from ~7-10 hours; BID dosing expected to reach targeted trough concentrations
Development Approach • Initial Phase 2 POC study to evaluate REC-3964 in combination with vancomycin • Focus on subjects at risk for CDI with moderate to severe disease planning to receive SOC therapy • Flexibility to assess effects of REC-3964 on both treatment and reduction of recurrence populations • Potential to generate early evidence of economic value and model cost-effectiveness of REC-3964 Vancomycin + REC-3964 Vancomycin Preliminary Phase 2 POC Design • Determination of optimal dose and sample size underway • Phase 2 initiation expected in 2024 Trial Update 43 Follow-up Enroll patients at high-risk for rCDI Planned Phase 2 Proof-of-Concept Trial Design Clinical: C. Difficile
RBM39: Novel CDK12-Adjacent Target for HR-Proficient Ovarian Cancer 44 RBM39 CDK12 2.5μM 1.0μM REC-65029 0.1μM 0.25μM CDK13 REC-65029 Similar Opposite BRCA-proficient ovarian cancer PDX Preclinical: HR-Proficient Ovarian Cancer Identify potential first-in-class tumor-targeted precision therapeutic NCE with novel MOA capable of potentially treating HR-proficient ovarian cancer Inhibition of target RBM39 (previously referred to as Target γ) may mimic the inhibition of CDK12 while mitigating toxicity related to CDK13 inhibition A Recursion-generated NCE showed single agent efficacy that is enhanced in combination with Niraparib in a BRCA-proficient PDX model IND-enabling studies are progressing GOAL INSIGHT FROM OS FURTHER CONFIDENCE NEXT STEPS Vehicle Niraparib REC-204 100 mpk REC-204 100 mpk + Niraparib OV0273 (PDX) in-vivo efficacy Survival data Note: in the OV0273 PDX model, mice were treated with a representative lead molecule REC-1170204 (100 mg/kg, BID, PO) ± Niraparib (40 mg/kg, QD, PO) for 32 days. Single agent REC-1170204 or in combination with Niraparib resulted in a statistically significant response vs either Niraparib or vehicle arms. In addition, there was a statistically significant improvement in survival > 30 days post final dose. *p<0.05, ** p<0.01, **** p<0.0001
Value driven by our team and our milestones
Advanced degreesEmployees Team Members ~550 What it takes to make this happen – a new kind of team and culture >50% ESG Highlights ✓ ESG reporting on Healthcare and Technology Metrics ✓ 100% of electricity powering our Biohive-1 supercomputer comes from renewable sources ✓ Learn more about Recursion’s ESG stewardship: www.recursion.com/esg ~43% Female Male ~56% ~1% Non-Binary Parity Pledge Signer gender parity and people of color parity Life Sciences – biology, chemistry, development, etc. Technology – data science, software engineering, automation, etc. Strategic Operations Community Impact Committed to ESG Excellence Founding Partner, Life Science Accelerator Founding Member, Life Science Collective 46 Data shown reflective of Q3 2023, gender statistics include participating individuals San Francisco, California Salt Lake City, Utah Toronto, Ontario Montréal, Québec
Our leadership team brings together experience & innovation to lead TechBio 47 Board of Directors Dean Li, MD PHD Co-Founder of RXRX, President of Merck Research Labs Rob Hershberg, MD PHD Co-Founder/CEO/Chairman of HilleVax, Former EVP/CSO/CBO of Celgene Blake Borgeson, PHD Co-Founder of RXRX Terry-Ann Burrell, MBA CFO & Treasurer, Beam Therapeutics Zavain Dar Co-Founder & Partner of Dimension R Martin Chavez, PHD Chairman of RXRX, Board Member of Alphabet, Vice-Chairman of 6th Street, Former CFO/CIO of GS Zachary Bogue, JD Co-Founder & Partner of Data Collective Chris Gibson, PHD Co-Founder & CEO STRICTLY CONFIDENTIAL Tina Larson President & COO Executive Team Ben Mabey Chief Technology Officer Kristen Rushton, MBA SVP of Business Operations Michael Secora, PHD Chief Financial Officer Chris Gibson, PHD Co-Founder & CEO Shafique Virani, MD FRCS Chief Business Officer Nathan Hatfield, JD MBA Chief Legal Officer Laura Schaevitz, PHD SVP and Head of Research David Mauro, MD PHD Chief Medical Officer Trademarks are the property of their respective owners and used for informational purposes only.
Strong Financial Position ~$387M in cash at end of Q3 2023 Near-Term • Potential option exercises for map building initiatives • Potential for additional partnership(s) in large, intractable areas of biology such as CV/Met • Potential additional option exercises for partnership programs • Ph2 initiation for AXIN1 or APC mutant cancers program expected in late Q4 2023 or early Q1 2024 • Ph2 initiation for C. difficile Infection program in 2024 • Potential to accelerate value creation with additional proprietary foundation models for biology (including patient data) and chemistry • Potential to open-source data and tools for non-commercial use and license data and tools to biopharma and other commercial users What to watch for at Recursion Medium-Term • Multiple Ph2 readouts for AI-discovered programs • CCM top-line data expected H2 2024 • NF2 & FAP safety & preliminary efficacy expected H2 2024 & H1 2025, respectively • Potential for additional INDs and clinical starts • Potential option exercises for partnership programs • Potential option exercises for map building initiatives or data sharing • Potential additional partnership(s) in large, intractable areas of biology and / or technological innovation • Recursion OS moves towards autonomous map building and navigation with digital and micro-synthetic chemistry Upcoming Potential Milestones 48Cash refers to cash and cash equivalents
Impact 49
Images adapted from Jayatunga, M., et al. Nature Reviews Drug Discovery 2022. The biopharmaceutical industry faces pressure amidst declining efficiency in drug discovery 50 Top-20 Pharma Companies AI-Native Drug Discovery Companies AI-enabled drug discovery efforts have proliferated alongside the declining efficiency of traditional approaches 705 768 708 602 575 641 686 709 519 439 393 333 0 500 1,000 1,500 2,000 2,500 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 Discovery / Preclinical Phase I Phase II Phase III 6 2 4 6 10 18 23 28 56 121 119 158 0 50 100 150 200 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 N u m b er o f C lin ic al a n d P re cl in ic al A ss e ts N u m b er o f C lin ic al a n d P re cl in ic al A ss e ts
Recursion’s map-based approach is designed to set the standard for drug discovery in the 21st century Platforms drive discovery. Unbiased & target agnostic Literature drives discovery. Informs target-based hypotheses Data are our fuel. Shape our hypotheses Data are an exhaust. Limited to testing hypotheses Virtuous cycles of atoms & bits. Iterative feedback accelerates learning Linear process. Little cross-program learning or iteration Connected data across programs. Relatable high-dimensional data Disparate data generation. Siloed to individual programs and diseases Industrialized to scale. Automation & standardization Bespoke processes. Low-dimensional assays & biomarkers 51
~50 different human cell types small molecule library, we believe this scale is on par with some large pharma companies hiPSC-derived cells produced since 2022, we believe that we are one of the largest hiPSC- derived cell producers ~1.7 Million ~1 Trillion Diverse Biological and Chemical Inputs Robotic Automation at Scale Up to 2.2 Million wet-lab experiments per week profiling genes and compounds, we believe we are one of the largest phenomics (human cellular image-based) data producers Digitization of Biology and Chemistry >25 Petabytes of proprietary high- dimensional data as of this filing, we believe this is one of the largest relatable in vitro biological and chemical datasets ML-Based Analysis Top 500 supercomputer across any industry (TOP500 List, Jun 2023), we leverage vast neural networks and multiomics approaches to extract features and drive insights relatable hypotheses across multiple biological and chemical contexts ML-Based RelationshipsHigh-Dimensional Validation 24K Up to 52 Top 500 >25 Petabytes2.2 Million / Week near whole exomes per week, we believe we are one of the largest transcriptomics data producers >5 Trillion Novel Insights at Scale Enables quality, relatability and scale of data Recursion OS Metrics shown reflective of Q3 2023 unless otherwise indicated
This is a whole-genome arrayed CRISPR knock-out Map generated in primary human endothelial cells Every gene is represented in a pairwise way (each is present in columns and rows) Dark Red indicates phenotypic similarity according to our neural networks while Dark Blue indicates phenotypic anti- similarity (which in our experience often suggests negative regulation) We can add the phenotypes of hundreds of thousands of small molecules at multiple doses and query and interact with these maps using a web application Thousands of examples of known biology and chemistry 53 All Human Genes with Significant Effects in this Cellular Context → A ll H u m an G en es w it h S ig n if ic an t Ef fe ct s in t h is C el lu la r C o n te xt → Genome-scale mapping
One such example – the JAK / STAT pathway clustered by strength of interaction, including both similar genes (red) and opposite genes (blue) Can wade into areas of novel biology and chemistry… 54 JAK1 TNKS1BP1 PPP1R9B PHF13 SOCS3 PRKCH MEGF8 ASB7 SLC39A1 DOCK9 ZMYM3 FAM49B STK24 YWHAB IL6ST IL6R STAT3 IL6 JA K 1 IL 6 ST A T3 IL 6 R IL 6 ST YW H A B ST K 2 4 FA M 4 9 B ZM YM 3 TN K S1 B P 1 P P P 1 R 9 B P H F1 3 SO C S3 P R K C H M EG F8 A SB 7 SL C 3 9 A 1 D O C K 9
Maps reveal known and novel biology Trademarks are the property of their respective owners and used for informational purposes only. 55 INTS9 INTS11 INTS4 INTS7 INTS2 INTS8 INTS1 INTS6 C7orf26 INTS10 INTS13 INTS14 IN TS 1 4 IN TS 1 3 IN TS 1 0 C 7 o rf 2 6 IN TS 6 IN TS 1 IN TS 8 IN TS 2 IN TS 7 IN TS 4 IN TS 1 1 IN TS 9 Phenomics TVN (below diagram) vs. Centerscale (above diagram) Similar Opposite • In 2022, new independent research identified a previously unknown gene, C7orf26, as part of the Integrator complex • Maps jointly developed by Recursion and Genentech replicated this same result • Demonstrates accuracy and consistency across different map building approaches
56 ~36 Billion Compounds from the Enamine Real Space ~80,000 predicted binding pockets from ~15,000 human proteins ~2.8 Quadrillion potential protein- ligand interactions computed and stored Recursion partnered with to integrate and optimize MatchMaker (acquired via ) for massive scale GPU-based computation on BioHive-1 and the DGXCloud This tool was deployed to predict protein- ligand interaction for ~36 Billion compounds from the Enamine Real Space, less than 90 days post-acquisition of Cyclica and less than 30 days post-partnership with NVIDIA Recursion will use the predicted interactions as a data-layer in its multi-omic dataset for honing mechanistic predictions from its wet- labs and for accelerating SAR cycles through better predictions for its internal pipeline and within its partnerships Bridging Protein and Chemical Space with Massive Protein-Ligand Interaction Predictions Computation at Scale Computation at Speed Computation as a Data-Layer
57
1 Includes approximately 500,000 compounds from Bayer’s proprietary library. 2 ‘Predicted Relationships’ refers to the number of Unique Perturbations that have been predicted using our maps. Biology and chemistry are complex – data that is reliable, relatable, and scalable is the Recursion differentiator 58 Year 2018 2019 2020 2021 2022 Total Phenomics Experiments (Millions) 8 24 56 115 175 Total Transcriptomics Experiments (Thousands) NA NA 2 91 258 Data (PB) 1.8 4.3 6.8 12.9 21.2 Cell Types 12 25 36 38 48 Unique Compounds Physically Housed at Recursion1 (Millions) 0.02 0.1 0.7 1.0 1.7 In Silico Chemistry Library (Billions) NA 0.02 3 12 >1,000 Predicted Biological and Chemical Relationships2 (Trillions) NA NA 0.01 0.2 3.1
• Recursion demonstrated that CRISPR-Cas9 editing induces chromosome arm-scale truncations across the genome • Creates a proximity bias in CRISPR screens which can confound some gene-gene relationships • Recursion demonstrated a correction method leveraging public CRISPR-Cas9 knockout screens to mitigate bias • Read “High-resolution genome-wide mapping of chromosome-arm-scale truncations induced by CRISPR-Cas9 editing” at www.biorxiv.org • Already in the top 5% of research outputs in online engagement www.altmetric.com CRISPR proximity bias revealed using genome-wide phenomics screens 59
Drug Prediction Correct? Hydroxychloroquine x ✓ Lopinavir x ✓ Ritonavir x ✓ Remdesivir ✓ ✓ Baricitinib ✓ ✓ Tofacitinib ✓ ✓ Fostamatinib ✓ ✓ Ivermectin* x ✓ Fluvoxamine x ✓ Dexamethasone x x COVID-19 research: Recursion OS correctly predicted 9 of 10 clinical trials 60https://www.biorxiv.org/content/10.1101/2020.04.21.054387v1 • Recursion conducted several AI-enabled experiments in April 2020 to investigate therapeutic potential for COVID-19 - Included FDA-approved drugs, EMA-approved drugs, and compounds in late-stage clinical trials for the modulation of the effect of SARS-CoV-2 on human cells • Experiments were compiled into the RxRx19 dataset (860+ GB of data) and made publicly available to accelerate the development of methods and pandemic treatments. * Recursion did not screen ivermectin, but did screen the related compounds selamectin and doramectin. Both of these tested negative; consequently ivermectin was not expected to have efficacy. Fostamatinib recently read out positive Ph3 results in COVID but was discontinued in ACTIV-4.
REC-994 for the Treatment of Symptomatic Cerebral Cavernous Malformations (CCM) Target / MOA Superoxide Scavenger Molecule Type Small Molecule Lead Indication(s) Cerebral Cavernous Malformations Status Phase 2 Designation(s) US & EU Orphan Drug Source of Insight Recursion OS
62 • Large unmet need for a novel nonsurgical treatment • Vascular malformations (cavernomas) in the brain and spinal cord • High-risk for hemorrhage creates “ticking time bomb” • Progressive increase in CCM size and number over time in those with familial disease • Debilitating symptoms, including intractable seizure, intracerebral hemorrhage, focal neurological deficits Description “Historically, cavernomas have been managed primarily with observation, surgical resection, and occasionally radiotherapy. However, for a number of reasons, many patients with cavernomas must endure a life with neurologic symptoms” - Ryan Kellogg, MD, Investigator at the University of Virginia Disease Overview : Cerebral Cavernous Malformations (CCM) 62 Clinical: CCM
63 Sources: Angioma Alliance ; Flemming KD, et al . Population-Based Prevalence of Cerebral Cavernous Malformations in Older Adults: Mayo Clinic Study of Aging. JAMA Neurol. 2017 Jul 1;74(7):801-805. doi: 10.1001/jamaneurol.2017.0439. PMID: 28492932; PMCID: PMC5647645 ; Spiegler S, et al Cerebral Cavernous Malformations: An Update on Prevalence, Molecular Genetic Analyses, and Genetic Counselling. Mol Syndromol. 2018 Feb;9(2):60-69. doi: 10.1159/000486292. Epub 2018 Jan 25. PMID: 29593473; PMCID: PMC5836221. ~360,000 Patient Population – Large and Diagnosable No Approved Medical Therapy • >1 million patients worldwide live with these lesions today • Caused by loss of function mutation in one of three genes: CCM1 (60%), CCM2 (20%), and CCM3 (20%) • Inherited autosomal dominant mutation in 30-40%; or sporadic • US symptomatic population is more than 5 times larger than other rare diseases like Cystic Fibrosis (>31k patients) and Spinal Muscular Atrophy (>33k patients) • No approved drugs for CCM • Most patients receive no treatment or only symptomatic therapy • Surgical resection or stereotactic radiosurgery not always feasible because of location of lesion and is not curative Julia – living with CCM Clinical: CCM Disease Overview : Cerebral Cavernous Malformations (CCM) Symptomatic US + EU5 patients 63
64 Disease Overview : CCM is an Under-Appreciated Orphan Disease Non-oncology Orphan Indication Product U.S. + EU5 Prevalence Cerebral cavernous malformation (CCM) REC-994 (Recursion) >1,800,000 (Symptomatic: ~360,000) Idiopathic pulmonary fibrosis (IPF) Esbriet (pirfenidone) >160,000 Cystic fibrosis (CF) VX-669/ VX-445 + Tezacaftor + Ivacaftor - Vertex >55,000 Spinal muscular atrophy (SMA) SPINRAZA (nusinersen) >65,000 Clinical: CCM Sources: Angioma Alliance ; Flemming KD, et al . Population-Based Prevalence of Cerebral Cavernous Malformations in Older Adults: Mayo Clinic Study of Aging. JAMA Neurol. 2017 Jul 1;74(7):801-805. doi: 10.1001/jamaneurol.2017.0439. PMID: 28492932; PMCID: PMC5647645 ; Spiegler S, et al Cerebral Cavernous Malformations: An Update on Prevalence, Molecular Genetic Analyses, and Genetic Counselling. Mol Syndromol. 2018 Feb;9(2):60-69. doi: 10.1159/000486292. Epub 2018 Jan 25. PMID: 29593473; PMCID: PMC5836221; Maher T, et al Global incidence and prevalence of idiopathic pulmonary fibrosis. Respir Res. 2021 Jul 7;22(197). Doi: 10.1186/s12931-021-01791-z. PMID: 34233665. DRG 2022 Solutions, Report: Epidemiology, Cystic Fibrosis. CDC: SMA 64
65 • Symptoms associated with both increased size of lesions, but also inflammation or activation of lesions within the immunoprivileged environment of the brain • Lesions arise from the capillary bed and are not high-pressure (e.g., the lesion growth is unlikely to be primarily driven by the law of Laplace) • The Recursion Vascular Stability Hypothesis: • Eliminating the lesions may not be required for significant patient benefit • Slowing or halting the growth of the lesions while mitigating lesion leakiness and endothelial cell activation to halt the feed-forward inflammatory reaction may mitigate some symptoms and be beneficial to patients Novel therapeutic approach Therapeutic Approach to Cerebral Cavernous Malformations (CCM) 65 Clinical: CCM
CCM – Applied prototyping of the Recursion OS Clinical: CCM Gibson, et al. Strategy for identifying repurposed drugs for the treatment of cerebral cavernous malformation. Circulation, 2015 siCTRL siCCM2 siCCM2 + Simvastatin siCCM2 + Cholecalciferol siCCM2 + REC-994 Using an early version of our Recursion OS in an academic setting, we identified about 39 molecules out of 2,100 screened that according to a machine learning classifier rescued a complex unbiased phenotype associated with CCM2 loss of function. Through a set of follow-on confirmatory assays of increasing complexity, REC-994 stood out as one of two compounds we tested in a 5-month chronic CCM animal model where both compounds demonstrated significant benefit. 66
Healthy REC-994 – Mechanism of Action REC-994 ImpactCCM Clinical: CCM • Endothelial cell activation • Smooth muscle proliferation • Leukocyte adhesion • Platelet aggregation By regulating SOD2, CCM1 (KRIT1) & CCM2 suppress: CCM1 or CCM2 loss of function leads to activated endothelium: • Decreased cell-cell junctional integrity and increased monolayer permeability • Impaired vasodilation • Cavernous angioma formation Dosing of REC-994 restores normal function: • Normalized ROS balance • Restores quiescent endothelial cell state • Stabilizes endothelial barrier function Adapted from REC-994 Investigator Brochure 67
68 Further Confidence : Preclinical Studies Confirm Insight Source: Data above from Gibson, et al. Strategy for identifying repurposed drugs for the treatment of cerebral cavernous malformation. Circulation, 2015 or Recursion internal data (Ccm1 mouse model) Preclinical Studies: REC-994 reduces lesion burden and ameliorates vascular defects in genetic mouse models of CCM Vascular permeability is a clinically relevant feature of CCM lesions REC-994 stabilizes the integrity of vasculature against challenges to permeability Clinical: CCM Reduces lesion number and size in Ccm1 and Ccm2 LOF mouse models1 Completely rescues acetylcholine-induced vasodilation defect2 Rescues dermal permeability defect in CCM2 mice3 Lesion size (mm2) Ccm1 LOF Model ecKO + REC-994 WT ecKO % V as o d ila ti o n Acetylcholine [Log M] 68 Lesion size (mm2) Ccm2 LOF Model * * * DMSO control REC-994 Ccm2 WT Ccm2 ecKO D e rm al P e rm e ab ili ty (A b so rp ti o n , A U ) *
Further Confidence : Clinical Studies Confirming Safety REC-994 Phase 1 Studies - well-tolerated with no dose-dependent adverse events in SAD and MAD Clinical: CCM MAD Study Placebo 50 mg 200 mg 400 mg 800 mg Total Number of TEAEs Total Subjects with ≥ one TEAE 5 4 0 0 10 3 4 3 15 4 Severity Mild Moderate Severe 3 1 0 0 0 0 3 0 0 3 0 0 3 1 0 Relationship to Study Drug None Unlikely Possibly Likely Definitely 3 1 0 0 0 0 0 0 0 0 0 1 0 2 0 2 1 0 0 0 1 2 0 1 0 Total Number of SAEs Total Subject with ≥ one TEAE Discontinued Study Drug Due to AE 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Source: REC-994 for the Treatment of Symptomatic Cerebral Cavernous Malformation (CCM) Phase 1 SAD and MAD Study Results. Oral Presentation at Alliance to Cure Scientific Meeting. 2022 Nov 17 69
REC-2282 for the Treatment of Progressive Neurofibromatosis Type 2 (NF2) Mutated Meningiomas Target / MOA HDAC Inhibitor Molecule Type Small Molecule Lead Indication(s) NF2 Mutated Meningiomas Status Phase 2/3 Designation(s) Fast Track; US and EU Orphan Drug Source of Insight Recursion OS
71 Disease Overview : Neurofibromatosis Type 2 (NF2) Source: https://rarediseases.org/rare-diseases/neurofibromatosis-2 Patient Population – Large and Diagnosable No Approved Medical Therapy • Rare autosomal dominant tumor syndrome resulting from biallelic inactivation of the NF2 gene which leads to deficiencies in the tumor suppressor protein merlin • NF2 can be inherited or spontaneous (>50% of patients represent new mutations); up to 1/3 are mosaic • CNS manifestations: meningiomas and vestibular schwannomas; mean age at presentation: ~20 years • No approved drugs for NF2 • Surgery is standard of care (when feasible) • Location may make complete resection untenable, leading to hearing loss, facial paralysis, poor balance and visual difficulty Ricki – living with NF2 Clinical: NF2 71
72 Disease Overview : Neurofibromatosis Type 2 (NF2) Meningiomas • Threatens mortality; if amenable, surgical excision is primary intervention • Many patients have multiple meningiomas that exhibit heterogenous behavior and asynchronous growth • Stasis or shrinkage of tumor could improve prognosis Clinical: NF2 ● Most tumors are benign and slow growing but location in CNS leads to serious morbidity or mortality ● Prognosis is adversely affected by early age at onset, a higher number of meningiomas and having a truncating mutation ~27,000 Patients who have Meningiomas that Harbor NF2 Mutations (Sporadic) ~6,000 NF2 Patients have Meningiomas (Familial) >66,000 Patients have Meningiomas Treatable US + EU5 patients ~33,000 Intracranial Meningioma Source: Pemov, et al. Comparative clinical and genomic analysis of neurofibromatosis type 2-associated cranial and spinal meningiomas. Nature. 2020 Jul 28;10(12563). Doi: https://doi.org/10.1038/s41598-020-69074-z; NORD 72
73 NF2 knockdown cells Healthy Cells REC-2282 REC-2282 identified as rescuing HUVEC cells treated with NF2 C o n tr o l N F2 s iR N A HUVEC, human umbilical vein endothelial cells; NF2, neurofibromatosis type 2; siRNA, small interfering RNA. Insight from OS : REC-2282 Rescued Loss of NF2 Clinical: NF2 73
74 REC-2282 – Mechanism of Action AKT, protein kinase B; eIF4F, eukaryotic initiation factor 4F; HDAC, histone deacetylase; mTor, mammalian target of rapamycin; mTORC1; mammalian target of rapamycin complex 1; NF2, neurofibromatosis type 2; PI3K, phosphoinositide 3-kinase; PP1, protein phosphate 1; Ras, reticular activating system. Clinical: NF2 Orally Bioavailable, CNS-penetrating, Small Molecule HDAC Inhibitor NF2 encodes for the protein Merlin and negatively regulates mTOR signaling 1 2 3 Loss of Merlin leads to increased signaling in the PI3K/AKT/mTOR pathway Oncogenic mTOR signaling arrested with HDAC inhibitors Cell proliferation and survival Normal cell proliferation and survival Cell proliferation and survival Constitutive activation is independent of extracellular factors and does not respond to biochemical signals that would normally regulate activity 74 1 2 3
75 REC-2282 preclinical studies demonstrated clear in-vivo efficacy in multiple NF2 tumor types Shrinks vestibular schwannoma xenografts in nude mice Prevents growth & regrowth of NF2- deficient meningioma model in mice Vehicle % Change tumor vol. REC-2282 % Change tumor vol. % c h an ge in t u m o r vo lu m e fr o m b as e lin e M R I % c h an ge in t u m o r vo lu m e fr o m b as e lin e M R I Further Confidence : Preclinical Studies Confirming Insight Clinical: NF2 https://link.springer.com/article/10.1007/s00280-020-04229-3 2 0% 10% 30% 60% -20% -40% 20% 40% 50% -10% -30% -50% 0% 10% -20% -40% -10% -30% -50% 75 1
• Evaluable Patients: CNS Solid Tumors: NF2 N=5; Non-CNS Solid Tumors: N=10 • PFS: CNS solid tumors = 9.1 months; Non-CNS solid tumors = 1.7 months • Best overall response = SD in 8/15 patients (53%; 95% CI 26.6–78.7) • Longest duration of follow-up without progression: > 27 months (N=1) • Most common AEs: cytopenia, fatigue, nausea Further Confidence : Prior Studies Suggest Potential Therapeutic Benefit 0 1 2 3 4 5 6 7 8 9 10 Overall Non-CNS solid tumors CNS solid tumors Months Progression-Free Survival 9.1m 1.7m 3.6m Clinical: NF2 Well understood clinical safety ... Multiple investigator-initiated studies in oncology indications Lengthy human clinical exposure in NF2 – multiple patients on drug for several years Well-characterized side effect profile … with a drug-like profile Established and scalable API manufacturing process Multiple cGMP batches of 10mg and 50mg tablets have been manufactured Excellent long-term stability 76
77 REC-2282 Appears Well Suited for NF2 vs Other HDAC Inhibitors 1 Sborov DW, et al. A phase 1 trial of the HDAC inhibitor AR-42 in patients with multiple myeloma and T- and B-cell lymphomas. Leuk Lymphoma. 2017 Oct;58(10):2310-2318. 2 Collier KA, et al. A phase 1 trial of the histone deacetylase inhibitor AR-42 in patients with neurofibromatosis type 2-associated tumors and advanced solid malignancies. Cancer Chemother Pharmacol. 2021 May;87(5):599-611. 3 Prescribing Information of Vorinostat/Belinostat/Romidepsin respectively Clinical: NF2 REC-2282 Would be First-In-Class HDAC Inhibitor for Treatment of NF2 Meningiomas 77
REC-4881 for the Treatment of Familial Adenomatous Polyposis (FAP) Target / MOA MEK Inhibitor Molecule Type Small Molecule Lead Indication(s) Familial Adenomatous Polyposis Status Phase 2 Designation(s) Fast Track; US and EU Orphan Drug Source of Insight Recursion OS
79 Disease Overview : Familial Adenomatous Polyposis Patient Population – Easily Identifiable Clinical: FAP ~50,000 Diagnosed US + EU5 patients • Autosomal dominant tumor predisposition syndrome caused by a mutation in the APC gene • Classic FAP (germline mutation) : • Hundreds to thousands of polyps in colon and upper GI tract • Extraintestinal manifestations (e.g., desmoid tumors) • 100% likelihood of developing colorectal cancer (CRC) before age 40, if untreated Polyps Found in Colon and Upper GI Tract 79 https://www.hopkinsmedicine.org/health/conditions-and-diseases/familial-adenomatous-polyposis
80 Disease Overview : Familial Adenomatous Polyposis – Standard of Care No Approved Medical Therapy • Standard of care: colectomy during adolescence (with or without removal of rectum) • Post-colectomy, patients still at significant risk of polyps progressing to GI cancer • Significant decrease in quality-of-life post-colectomy: continued endoscopies and surgical intervention Polyps on mucosal membrane of colon Cross section of colon and rectum Multiple polyps in the colon Sigmoidoscope Scope view Clinical: FAP “Despite progress with surgical management, the need for effective therapies for FAP remains high due to continued risk of tumors post-surgery” - Niloy Jewel Samadder, MD, Mayo Clinic https://www.hopkinsmedicine.org/health/conditions-and-diseases/familial-adenomatous-polyposis 80
81 REC-4881 rescued phenotypic defects of cells with APC knockdown 0.1 µM REC-4881 Insight from OS : Rescued Loss of APC, Inhibited Tumor Growth • Compared to thousands of other molecules tested, REC-4881 rescued phenotypic defects substantially better (including better rescue than other MEK inhibitors) for APC specific knockdown • Findings validated in tumor cell lines and spheroids grown from human epithelial tumor cells with APC mutation • 1,000x more selectivity in tumor cell lines with APC mutation • Inhibited growth and organization of spheroids APC knockdown cellsHealthy Cells Clinical: FAP 81
82 Jeon, WJ, et al. (2018). Interaction between Wnt/β-catenin and RAS-ERK pathways and an anti-cancer strategy via degradations of β-catenin and RAS by targeting the Wnt/β-catenin pathway. npj Precision Oncology, 2(5). 3 3 REC-4881 inhibits MEK 1/2 and recovers the destabilization of RAS by the β-Catenin destruction complex, restoring the cell back to a Wnt-off like state 2 1 Orally Bioavailable, Small Molecule MEK Inhibitor Disease State REC-4881 Impact MoA : REC-4881 Blocks Wnt Mutation Induced MAPK Signaling Clinical: FAP 82
83 Further Confidence : Preclinical Studies Confirming Reduction in Polyp Count and High-Grade Dysplasia APC, adenomatosis polyposis coli; ERK, extracellular signal-regulated kinase; FAP, familial adenomatous polyposis. Clinical: FAP ↓ High-Grade Dysplasia 2• In-vivo efficacy in APCmin mouse model • Apcmin = FAP disease model • Mice treated once daily for 8 weeks After 8 weeks of treatment: ↓ Polyp Count1 1 2 To ta l P o ly p C o u n t (+ /- S EM ) H ig h G ra d e A d e n o m as ( % ) 83
84 Note: AE, adverse event; MEK, mitogen-activated protein kinase; NHV, normal healthy volunteer; pERK, phosphorylated extracellular signal-regulated kinase; SAE, serious adverse event. REC-4881-101: Single-center, double-blind, placebo- controlled, dose-escalation study in healthy volunteers • Group 1 (n=13): Food effect crossover (REC-4881 4 mg/PBO [fed/fasted]), followed by single dose REC-4881 8 mg/PBO [fed] • Group 2 (n=12): Matched single ascending dose (REC- 4881 4 mg/PBO; REC-4881 8 mg/PBO; REC-4881 12 mg/PBO) Accomplished Further Confidence : Clinical Data Generated by Recursion Clinical: FAP Recursion formulation yields exposures comparable to Takeda’s formulation (molecule in-licensed from Takeda) No food effect Dose proportional increases in exposure Similar to C20001 study, observed pERK inhibition (i.e., target engagement) at 8 mg and 12 mg doses Acceptable safety profile 84
REC-4881 for the Treatment of Solid Tumors with AXIN1 or APC Mutant Cancers Target / MOA MEK Inhibitor Molecule Type Small Molecule Lead Indication(s) Solid Tumors with AXIN1 or APC Mutant Cancers Status Phase 2 Source of Insight Recursion OS
86 Disease Overview : AXIN1 or APC Mutant Cancers • Sustained Wnt signaling is a frequent driver event found across a wide variety of solid tumors • Dysregulation of β-catenin destruction complex due to inactivating mutations in AXIN1 or APC leads to sustained Wnt signaling promoting cancer progression and survival1 • AXIN1 or APC mutant solid tumors are considered clinically aggressive and resistant to standard treatments 1 Bugter, J.M., et al. Nat Rev Cancer, 2021, 21, pp.5-21 Gross morphology of HCC tumor Clinical: AXIN1 or APC “Nothing in HCC has immediate therapeutic relevance and the most common mutations are TERT, TP53, and Wnt (CTNNB1/AXIN1/APC) and combined these alterations define almost 80% of patients and are not targetable” - KOL, Clinical Investigator, Texas 86
87 • AXIN1 and APC genes covered by commercially available NGS panels and liquid biopsy detection assays • FDA guidance supports utility of ctDNA as patient selection for the detection of alterations for eligibility criteria and as a stratification factor for trials enrolling marker-positive and marker-negative populations3 • Multiple tumor types will inform study design and patient selection Flexible Patient Selection Strategy and Study Design 1 Obtained from cbioportal.org. 2 Represents 2L treatable population estimates; obtained from DRG. 3 https://www.fda.gov/media/158072/download Tumor Type AXIN1 Mutation Frequency1 APC Mutation Frequency1 Treatable Population2 (US+EU5) CRC 3% 70% 27,450 LUAD 4% 11% 14,000 Prostate 2% 11% 6,700 Bladder 3% 8% 5,100 HCC 12% 5% 3,100 Endometrial 8% 12% 2,600 Esophageal 2% 7% 2,600 PDAC 1% 2% 1,500 Ovarian 1% 3% 1,400 TNBC 1% 2% 300 Preclinical data with REC-4881 at clinically relevant exposures in HCC and Ovarian PDX mouse models gives confidence to pursue other mutant cancer types Clinical: AXIN1 or APC Disease Overview : AXIN1 or APC Mutant Cancers 87 ~65,000
REC-4881 Dosage Hypothesis: Rescue of AXIN1 may impact tumor progression and/or restore checkpoint sensitivity in cancers driven by AXIN1 loss Recursion Differentiation: REC-4881 rescues tumor suppressor genes APC and AXIN1 • APC and AXIN1 are negative regulators of Wnt signaling • Both proteins form part of the B-catenin destruction complex. Strong clustering suggests map recapitulation of this biology Insight from OS : Novel Insight around Established MoA Clinical: AXIN1 or APC Heat map from Recursion OS Similar Opposite 88
Further Confidence : Preclinical Studies Confirming Insight Clinical: AXIN1 or APC Note: REC-4881 dosed at 3 mg/kg QD for up to 21 days. 3 mice per treatment per model (3 x 3 x 3) design. 1 Wong, H., et al. Clin Cancer Res, 2012, 18:14, pp.3846-3855 89 -100 -80 -60 -40 -20 0 20 AXIN1 or APC wildtype AXIN1 or APC mutant 0 5 10 15 20 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Wildtype Time (days elapsed) P ro b a b il it y o f p ro g re s s io n f re e (b y t u m o r d o u b li n g ) Vehicle (n = 52) REC-4881 (n = 45) Median PFS (days) 95% CI 7.0 9.0 (4.70 - 10.43) (6.04 - 13.41) Log-rank p value = 0.23 HR = 0.81 (95% CI 0.55 - 1.21) b 0 5 10 15 20 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Mutant Time (days elapsed) P ro b a b il it y o f p ro g re s s io n f re e (b y t u m o r d o u b li n g ) Vehicle (n = 33) REC-4881 (n = 33) Median PFS (days) 95% CI 7.0 12.0 (4.19 - 11.70) (7.18 - 20.01) Log-rank p value < 0.001 HR = 0.49 (95% CI 0.29 - 0.83) a • Significantly greater antitumor activity observed with REC- 4881 in mutant models versus wildtype • Majority of mutant models ≥ 60% tumor growth inhibition, which is considered a benchmark for a response in the clinic1 Average Response : ~70% mutant vs ~49% wildtype Median Response: ~72% mutant vs ~48% wildtype Efficacy found in In Vivo Mice Models … 0 5 10 15 20 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Wildtype Time (days elapsed) P ro b a b il it y o f p ro g re s s io n f re e (b y t u m o r d o u b li n g ) Vehicle (n = 52) REC-4881 (n = 45) Median PFS (days) 95% CI 7.0 9.0 (4.70 - 10.43) (6.04 - 13.41) Log-rank p value = 0.23 HR = 0.81 (95% CI 0.55 - 1.21) b 0 5 10 15 20 0. 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Mutant Time (days elapsed) P ro b a b il it y o f p ro g re s s io n f re e (b y t u m o r d o u b li n g ) Vehicle (n = 33) REC-4881 (n = 33) Median PFS (days) 95% CI 7.0 12.0 (4.19 - 11.70) (7.18 - 20.01) Log-rank p value < 0.001 HR = 0.49 (95% CI 0.29 - 0.83) a Tu m o r G ro w th (% C h an ge v s V eh ic le ) … Led to Significant Progression Free Survival
REC-3964 for the Treatment of C. Difficile Infection Target / MOA Selective C. diff Toxin Inhibitor Molecule Type Small Molecule Lead Indication(s) C. Difficile Infection Status Phase 2 Source of Insight Recursion OS
Source, CDC *NAAT = Nucleic Acid Amplification Test; **rCDI = recurrent CDI • RCDI** occurs in 20-30% of patients treated with standard of care • 40% of those patients will continue to recur with 2+ episodes • >29,000 patients die in the US each year from CDI • Cost burden of up to $4.8bn annually 91 Disease Overview : C. Difficile Infection (CDI) Patient Population – Large, Diagnosable and Easy to Identify Large, Unmet Need with Significant Cost Burden • Symptoms caused by clostridioides difficile tissue-damaging toxins released in the colon • Patients who experience >3 unformed stools are diagnosed via NAAT* for toxin gene or positive stool test for toxins • Patients who are at highest risk are those on antibiotics, and frequently visit hospitals or are living in a nursing home • More than 80% of cases occur among patients age 65 or older Clinical: C. Difficile ~730,000 Diagnosed US + EU5 patients Colleen – lived with rCDI 91
92 Disease Overview : C. Difficile Infection (CDI) Disruption of microbiota and colonization of C. diff Release of C. diff toxins Degradation of colon cell junction & toxin transit to bloodstream Clinical: C. Difficile Source: McCollum, D,, Rodriguez, JM . Detection, Treatment, and Prevention of Clostridium difficile Infection. Clinical Gastroenterology and Hepatology 2012 Mar 19. https://doi.org/10.1016/j.cgh.2012.03.008 92 1 2 3
REC-3964 identified as a NCE that demonstrated strong rescue in HUVEC cells treated with C. diff toxin Insight from OS : REC-3964 Rescued Cells Treated with C. Difficile Toxins C. diff toxin B phenotype Healthy Control Disease State Healthy Cells REC-3964 0.1 µM Clinical: C. Difficile 93
REC-3964 : Selective Inhibitor of C. Difficile Toxins Clinical: C. Difficile The glucosyltransferase locks Rho family GTPases in the inactive state C.diff toxins bind to cell surface receptors and trigger endocytic event 1 Autocatalytic cleavage event releases C.diff toxin's glucoyltransferase enzymatic domain into the cytosol of the infected cell 2 3 3 1 2 Inactivation of Rho GTPases alters cytoskeletal dynamics, induces apoptosis, and impairs barrier function which drives the pathological effects of C.diff infection 4 4 REC-3964 is Recursion’s 1st Small Molecule NCE to Reach the Clinic 94Adapted from Awad et al. 2014
Adapted from Awad, MM. et al. (2014). Clostridium difficile virulence factors: Insights into an anaerobic spore-forming pathogen. Gut Microbes. 5(5), 579-593. REC-3964 : Selective Inhibitor of C. Difficile Toxins Clinical: C. Difficile 3 1 2 4 REC-3964 binds and blocks catalytic activity of the toxin’s innate glucosyltransferase, but not the host's 5 5 REC-3964 is Recursion’s 1st Small Molecule NCE to Reach the Clinic 95
Further Confidence : Preclinical Studies Confirmed Recursion OS Insight ✓ REC-3964 restores gut epithelial barrier integrity, which when disrupted causes inflammation and diarrhea REC-3964 rescues barrier integrity with increasing concentrations REC-3964 improved probability of survival in a hamster model of C. difficile infection ✓ Improved probability of survival beyond treatment completion Clinical: C. Difficile Healthy Monolayer (100%) Toxin-Damaged Monolayer (0%) REC-3964 Concentration (Log μM) R e si st an ce ( n o rm al iz e d % ) 96
MAD Study Placebo (N=8) n ( %) 100 mg (N=10) n ( %) 300 mg (N=8) n ( %) 500 mg (N=8) n ( %) 900 mg (N=8) n ( %) REC-3964 Overall (N=34) n ( %) MAD Overall (N=42) n ( %) Total Number of TEAEs 17 24 5 9 7 45 62 Total Participants with ≥ 1 TEAE 6 ( 75.0) 8 ( 80.0) 4 ( 50.0) 5 ( 62.5) 4 ( 50.0) 21 ( 61.8) 27 ( 64.3) Relationship to Study Drug Not Related 4 ( 50.0) 6 ( 60.0) 3 ( 37.5) 4 ( 50.0) 4 ( 50.0) 17 ( 50.0) 21 ( 50.0) Related 2 ( 25.0) 2 ( 20.0) 1 ( 12.5) 1 ( 12.5) 0 4 ( 11.8) 6 ( 14.3) Abdominal Distension 2 ( 25.0) 1 ( 10.0) 1 ( 12.5) 1 ( 12.5) 0 3 ( 8.8) 5 ( 11.9) Flatulence 0 1 ( 10.0) 0 0 0 1 ( 2.9) 1 ( 2.4) Severity Grade 1 6 ( 75.0) 8 ( 80.0) 4 ( 50.0) 5 ( 62.5) 4 ( 50.0) 21 ( 61.8) 27 ( 64.3) Grade 2 0 0 0 0 0 0 0 Grade ≥ 3 0 0 0 0 0 0 0 Total Number of SAEs 0 0 0 0 0 0 0 Discontinued Study Drug Due to AE 0 0 0 0 0 0 0 Further Confidence : Clinical Studies Confirming Safety REC-3964 was well-tolerated with no treatment-related SAEs 97 Clinical: C. Difficile TEAEs = treatment emergent adverse events; Grade 1 = Mild, Grade 2 = Moderate, Grade 3 = Severe, Grade 4 = Life Threatening, Grade 5 = Fatal
