10-K

rlay-10k_20201231.htm

10-K

rlay-10k_20201231.htm

UNITED STATES

SECURITIES AND EXCHANGE COMMISSION

Washington, D.C. 20549

FORM 10-K

(Mark One)

For the fiscal year ended December 31, 2020

Commission File Number: 001-39385

RELAY THERAPEUTICS, INC.

(Exact name of Registrant as specified in its Charter)

399 Binney Street, 2nd Floor Cambridge, MA 02139

(Address of principal executive offices) (Zip Code)

Registrant’s telephone number, including area code: (617) 370-8837

Securities registered pursuant to Section 12(b) of the Act:

Title of each class Trading Symbol(s) Name of each exchange on which registered

Common Stock, par value $0.001 per share RLAY Nasdaq Global Market

Securities registered pursuant to Section 12(g) of the Act: None

Indicate by check mark if the Registrant is a well-known seasoned issuer, as defined in Rule 405 of the Securities Act. Yes☐ No☒

Indicate by check mark if the Registrant is not required to file reports pursuant to Section 13 or 15(d) of the Act. Yes☐No☒

Indicate by check mark whether the Registrant: (1) has filed all reports required to be filed by Section 13 or 15(d) of the Securities Exchange Act of 1934 during the preceding 12 months (or for such shorter period that the Registrant was required to file such reports), and (2) has been subject to such filing requirements for the past 90 days. Yes☒No☐

Indicate by check mark whether the Registrant has submitted electronically every Interactive Data File required to be submitted pursuant to Rule 405 of Regulation S-T (§232.405 of this chapter) during the preceding 12 months (or for such shorter period that the Registrant was required to submit such files). Yes☒No☐

Indicate by check mark whether the registrant is a large accelerated filer, an accelerated filer, a non-accelerated filer, smaller reporting company, or an emerging growth company. See the definitions of “large accelerated filer,” “accelerated filer,” “smaller reporting company,” and “emerging growth company” in Rule 12b-2 of the Exchange Act.

Large accelerated filer ☐ Accelerated filer ☐

Non-accelerated filer ☒ Smaller reporting company ☒

Emerging growth company ☒

If an emerging growth company, indicate by check mark if the registrant has elected not to use the extended transition period for complying with any new or revised financial accounting standards provided pursuant to Section 13(a) of the Exchange Act. ☐

Indicate by check mark whether the registrant has filed a report on and attestation to its management’s assessment of the effectiveness of its internal control over financial reporting under Section 404(b) of the Sarbanes-Oxley Act (15 U.S.C. 7262(b)) by the registered public accounting firm that prepared or issued its audit report. ☐

Indicate by check mark whether the Registrant is a shell company (as defined in Rule 12b-2 of the Exchange Act). Yes☐ No☒

The aggregate market value of the Registrant’s common stock held by non-affiliates of the Registrant was $2,543,277,454 as of the closing of the Registrant’s initial public offering on July 20, 2020 (based on a closing price of $42.87 per share as quoted by the Nasdaq Global Market as of such date). In determining the market value of non-affiliate common stock, shares of the Registrant’s common stock beneficially owned by officers, directors and affiliates have been excluded. This determination of affiliate status is not necessarily a conclusive determination for other purposes.

The number of shares of Registrant’s Common Stock outstanding as of March 23, 2021 was 90,399,972.

DOCUMENTS INCORPORATED BY REFERENCE

The registrant intends to file a definitive proxy statement pursuant to Regulation 14A relating to the 2021 Annual Meeting of Stockholders within 120 days of the end of the registrant’s fiscal year ended December 31, 2020. Portions of such definitive proxy statement are incorporated by reference into Part III of this Annual Report on Form 10-K to the extent stated herein.

Table of Contents

Page

PART I

Item 1. Business 5

Item 1A. Risk Factors 67

Item 1B. Unresolved Staff Comments 118

Item 2. Properties 118

Item 3. Legal Proceedings 118

Item 4. Mine Safety Disclosures 118

PART II

Item 7A. Quantitative and Qualitative Disclosures About Market Risk 133

Item 8. Financial Statements and Supplementary Data 134

Item 9A. Controls and Procedures 134

Item 9B. Other Information 135

PART III

Item 10. Directors, Executive Officers and Corporate Governance 136

Item 11. Executive Compensation 136

Item 14. Principal Accounting Fees and Services 136

PART IV

Item 15. Exhibits, Financial Statement Schedules 137

Summary of the Material Risks Associated with Our Business

SPECIAL NOTE REGARDING FORWARD-LOOKING STATEMENTS

This Annual Report on Form 10-K contains express or implied forward-looking statements that are based on our management’s belief and assumptions and on information currently available to our management. Although we believe that the expectations reflected in these forward-looking statements are reasonable, these statements relate to future events or our future operational or financial performance, and involve known and unknown risks, uncertainties and other factors that may cause our actual results, performance or achievements to be materially different from any future results, performance or achievements expressed or implied by these forward-looking statements. Forward-looking statements contained in this Annual Report on Form 10-K include, but are not limited to, statements about:

• our ability to commercialize our products, if approved;

• the pricing and reimbursement of our product candidates, if approved;

• our financial performance;

• the rate and degree of market acceptance of our product candidates;

• regulatory developments in the United States and foreign countries;

• the success of competing therapies that are or may become available;

• our ability to attract and retain key scientific or management personnel;

• the impact of laws and regulations;

• developments relating to our competitors and our industry;

In some cases, you can identify forward-looking statements by terminology such as “may,” “should,” “expects,” “intends,” “plans,” “anticipates,” “believes,” “estimates,” “predicts,” “potential,” “continue” or the negative of these terms or other comparable terminology. These statements are only predictions. You should not place undue reliance on forward-looking statements because they involve known and unknown risks, uncertainties, and other factors, which are, in some cases, beyond our control and which could materially affect results. Factors that may cause actual results to differ materially from current expectations include, among other things, those listed above under “Summary of the Material Risks Associated with Our Business” and under the section titled “Risk Factors” and elsewhere in this Annual Report on Form 10-K. If one or more of these risks or uncertainties occur, or if our underlying assumptions prove to be incorrect, actual events or results may vary significantly from those implied or projected by the forward-looking statements. No forward-looking statement is a guarantee of future performance. You should read this Annual Report on Form 10-K and the documents that we reference in this Annual Report on Form 10-K and have filed with the Securities and Exchange Commission, or the SEC, as exhibits hereto completely and with the understanding that our actual future results may be materially different from any future results expressed or implied by these forward-looking statements.

The forward-looking statements in this Annual Report on Form 10-K represent our views as of the date of this Annual Report on Form 10-K. We anticipate that subsequent events and developments will cause our views to change. However, while we may elect to update these forward-looking statements at some point in the future, we have no current intention of doing so except to the extent required by applicable law. You should therefore not rely on these forward-looking statements as representing our views as of any date subsequent to the date of this Annual Report on Form 10-K.

This Annual Report on Form 10-K also contains estimates, projections and other information concerning our industry, our business and the markets for our product candidates. Information that is based on estimates, forecasts, projections, market research or similar methodologies is inherently subject to uncertainties and actual events or circumstances may differ materially from events and circumstances that are assumed in this information. Unless otherwise expressly stated, we obtained this industry, business, market, and other data from our own internal estimates and research as well as from reports, research surveys, studies, and similar data prepared by market research firms and other third parties, industry, medical and general publications, government data and similar sources. While we are not aware of any misstatements regarding any third-party information presented in this Annual Report on Form 10-K, their estimates, in particular as they relate to projections, involve numerous assumptions, are subject to risks and uncertainties and are subject to change based on various factors, including those discussed under the section titled “Risk Factors” and elsewhere in this Annual Report on Form 10-K.

PART I

Except where the context otherwise requires or where otherwise indicated, the terms “Relay Therapeutics,” “we,” “us,” “our,” “our company,” the “Company,” and “our business” refer to Relay Therapeutics, Inc. and its consolidated subsidiary.

Item 1. Business.

Overview

We are a clinical-stage precision medicines company transforming the drug discovery process with an initial focus on enhancing small molecule therapeutic discovery in targeted oncology. Our company is built upon unparalleled insights into protein motion and how this dynamic behavior relates to protein function. These insights may enable us to more effectively drug protein targets that previously have been intractable (i.e., inadequately drugged or undruggable). We believe we have a differentiated approach to drug these protein targets based on their motion, which enables us to select and advance unique product candidates. We built our DynamoTM platform to integrate an array of leading edge experimental and computational approaches, which allows us to apply our understanding of protein structure and motion to drug discovery.

We are advancing a pipeline of medicines to address targets in precision oncology, including our lead product candidates, RLY-1971 and RLY-4008, as well as our PI3Kα mutant selective program, or the RLY-PI3K1047 program. We initiated a Phase 1 clinical trial for RLY-1971, our inhibitor of Src homology region 2 domain-containing phosphatase-2, or SHP2, in patients with advanced solid tumors in the first quarter of 2020. In December 2020, we entered into a global collaboration and license agreement, or the Genentech Agreement, with Genentech, Inc., a member of the Roche Group, or Genentech, for the development and commercialization of RLY-1971. We initiated a first-in-human clinical trial of RLY-4008, our inhibitor of fibroblast growth factor receptor 2, or FGFR2, enriched for patients with advanced solid tumors having oncogenic FGFR2 alterations in the third quarter of 2020. We anticipate the RLY-PI3K1047 program, our program for molecules targeting cancer-associated mutant variants of phosphoinostide 3-kinase alpha, or PI3Kα, to be in Investigational New Drug, or IND, enabling studies in 2021. While our initial focus is on precision oncology, we believe our Dynamo platform may also be broadly applied to other areas of precision medicine, such as genetic disease. In addition to the three product candidates described above, we have five discovery stage programs across precision oncology and genetic disease. We are focused on using the novel insights derived from our approach to transform the lives of patients suffering from debilitating and life-threatening diseases through the discovery, development and commercialization of our therapies.

Precision medicine emerged as an approach for disease treatment as the understanding of the link between genetic alterations, protein dysfunction and diseases evolved. Precision medicine aims to specifically and potently drug genetically validated target proteins (i.e., genetic variants potentially implicated in biology of disease). However, some target proteins thus far have been intractable using conventional drug discovery tools, such as structure-based drug design, or SBDD. While SBDD is well-suited to solving some drug discovery problems such as orthosteric site kinase inhibitors, its reliance on static images of protein fragments limits its ability to gain accurate insights into the dynamic behavior of proteins in their natural state, which in turn limits its ability to discover medicines with exquisite specificity. Our approach pivots the understanding of protein targets from the industry-standard, static view, to a novel paradigm based on fundamental insights into protein motion. We then apply these novel insights into protein motion to drug discovery and design, which we term Motion Based Drug DesignTM, or MBDD.

The confluence of three forces — the proliferation of readily available genomic data, the evolution of experimental techniques, and advancements in computational power and speed — led to the founding of Relay Therapeutics. We believe we are uniquely situated in our ability to consolidate these advances and, when combined with our world-class team of both experimental and computational experts, integrate these solutions into MBDD to create medicines that will make a transformative difference for patients.

Key Drug Discovery Steps of Our Dynamo Platform

Our Dynamo platform puts protein motion at the center of drug discovery and design, integrating a broad and tailored array of leading-edge experimental and computational approaches, including deploying the Anton 2

supercomputer, which was custom-built by D. E. Shaw Research, LLC, or D. E. Shaw Research, to perform molecular dynamic simulations of proteins. We have access to the Anton 2 supercomputer, which we believe to be the only resource of computational power of its caliber, through our collaboration with D. E. Shaw Research, LLC, or D.E. Shaw Research, pursuant to which we collaborate with D. E. Shaw Research to rapidly develop various protein models. Our use of the Anton 2 supercomputer and our collaboration with D.E. Shaw Research is subject to the terms and conditions of the Amended and Restated Collaboration and License Agreement with D. E. Shaw Research, or the DESRES Agreement. See “—Our Collaborations—License Agreements and Strategic Collaborations—Collaboration and License Agreement with D. E. Shaw Research, LLC.” We deploy the power of the platform in three key phases of MBDD discovery:

Our Dynamo platform has the potential to address a diverse range of disease targets, including those proteins that have not been addressed selectively and potently with existing therapies. While we have initially focused our Dynamo platform on small molecule drug discovery in the area of precision oncology, we believe it could be readily deployed across broader precision and genetic medicine areas as well as other therapeutic modalities, such as protein therapeutics and antibody design.

Our Programs

We have deployed our technology platform to build a pipeline of product candidates to address targets in precision oncology, where there is clear evidence linking target proteins to disease and where molecular diagnostics can unambiguously identify relevant patients for treatment. We believe this approach will increase the likelihood of successfully translating a specific pharmacological mechanism into clinical benefit. The targets associated with all of our current programs are Category 1 Targets under our DESRES Agreement. See “—Our Collaborations—License Agreements and Strategic Collaborations —Collaboration and License Agreement with D. E. Shaw Research, LLC.”

RLY-4008

RLY-4008 is designed to be an oral, small molecule, selective inhibitor of fibroblast growth factor receptor 2, or FGFR2, a receptor tyrosine kinase that is frequently altered in certain cancers. FGFR2 is one of four members of the FGFR family, a set of closely related proteins with highly similar protein sequences and properties. RLY-4008 demonstrates FGFR2-dependent killing in cancer cell lines, while showing minimal inhibition of other targets, including other members of the FGFR family. We initiated a first-in-human clinical trial of RLY-4008 enriched for patients with advanced solid tumors having oncogenic FGFR2 alterations in September 2020. We anticipate giving an initial clinical update on this trial in the second half of 2021. We believe FGFR2-mediated cancers affect approximately 8,000 late-line patients annually in the United States. In the future, if RLY-4008 advances to earlier lines of treatment, we believe it could potentially address approximately 20,000 patients annually in the United States.

Mutant-PI3Kα Inhibitor Program

RLY-PI3K1047 is a lead compound in our franchise of programs targeting cancer-associated mutant variants of phosphoinositide 3-kinase alpha, or PI3Kα. RLY-PI3K1047 is a small molecule inhibitor of PI3Kα that we designed specifically to target PI3Kα H1047X mutants via a previously undescribed allosteric mechanism. Oral dosing of RLY-PI3K1047 resulted in tumor growth inhibition in mouse xenograft models of PI3Kα H1047R mutant carcinoma. We expect to begin IND-enabling studies for a differentiated PI3Kα H1047X mutant-selective inhibitor in 2021. We believe PI3Kα H1047X mutant cancers affect approximately 10,000 late-line patients annually in the United States. In the future, if RLY-PI3K1047 advances to earlier lines of treatment, we believe it could potentially be suitable for use in approximately 50,000 patients annually in the United States.

Two additional mutations of interest for our PI3Kα franchise are E542X and E545X. We estimate there are approximately 15,000 late-line and 60,000 total patients annually in the United States who might benefit from a PI3Kα targeted inhibitor that targets the mutations at E542 and E545.

RLY-1971

RLY-1971 binds and stabilizes SHP2 in its inactive conformation. SHP2 promotes cancer cell survival and growth through the RAS pathway by transducing signals downstream from receptor tyrosine kinases, or RTKs. Additionally, activating SHP2 mutations causes enhanced signaling in the absence of ligand stimulation and has been identified as an oncogenic driver in a range of tumors. As a critical signaling node and regulator, SHP2 drives cancer cell proliferation and plays a key role in the way cancer cells develop resistance to targeted therapies. We believe that inhibition of SHP2 could be effective as a monotherapy in cancers with specific alterations and could block a common path that cancer cells exploit to resist other antitumor agents, thus overcoming or delaying the onset of resistance to those therapies.

We are currently evaluating the safety and tolerability of RLY-1971 in a Phase 1 dose escalation study in patients with advanced or metastatic solid tumors. In December 2020, we entered into the Genentech Agreement, a global collaboration and license agreement with Genentech for the development and commercialization of RLY-1971. Future development for RLY-1971 will be governed by a joint development team between us and Genentech. We expect a combination trial of RLY-1971 and Genentech’s KRASG12C inhibitor, GDC-6036, to be initiated in 2021. Given the range of cancers that are related to SHP2 dependence, we believe RLY-1971 could serve as a backbone for compelling combination therapies. We believe SHP2-mediated cancers affect approximately 55,000 late-line patients annually in combination therapy settings in the United States. In the future, if RLY-1971 advances to earlier lines of treatment, we believe it could potentially have applicability to approximately 90,000 patients annually in the United States.

Under the terms of the Genentech Agreement, we have received $75 million in an upfront payment and are eligible to receive $25 million in near-term payments; and, if we do not opt into a U.S. profit/cost share, up to $695 million in additional development, commercialization and sales-based milestones for RLY-1971; and tiered royalties on annual global net sales (on a country-by-country basis), in the low-to-mid-teens, subject to reduction in certain circumstances. Additionally, we are eligible to receive additional royalties in the event of regulatory approval of RLY-1971 and Genentech’s compound, GDC-6036, that directly binds to and inhibits KRASG12C, in combination. We have the right to opt-in to a 50/50 U.S. profit/cost share and if we do opt into the U.S. profit/cost share, we are eligible to receive up to $410 million in additional commercialization and sales-based milestones for RLY-1971 outside of the U.S. and tiered royalties on annual net sales outside of the U.S. (on a country-by-country basis), in the low-to-mid-teens, subject to reduction in certain circumstances. We also retain the right to develop RLY-1971 in combination with our FGFR2 and PI3Kα programs. If we elect to opt-out of the profit/cost share, then the milestone and royalty payment obligations will revert to the financial terms that would be applicable if we had not opted into the profit/cost share, with certain adjustments. See “—Our Collaborations—License Agreements and Strategic Collaborations—Genentech Collaboration and License Agreement” for more details on the Genentech Agreement.

Discovery Programs

We are deploying our Dynamo platform and MBDD approach to advance multiple discovery-stage precision oncology programs. As with our lead programs, these programs leverage insights into protein conformational dynamics to address high-value, genetically validated oncogenes that previously have been intractable to conventional drug-discovery approaches. Our Dynamo platform’s protein visualization capabilities can be applied to multiple therapeutic areas beyond precision oncology. To further diversify our pipeline, we are leveraging our Dynamo platform to address validated targets in monogenic diseases, where genetic alterations lead to disease-causing defects in protein motion.

Our Strategy

Our mission is to leverage unique insights into protein motion to transform the lives of patients suffering from debilitating and life-threatening diseases through the discovery, development and commercialization of small molecule therapies. We believe that, by placing protein motion at the heart of MBDD discovery, our unique Dynamo platform has the potential to address previously intractable precision medicine targets. To accomplish this, we intend to continue building a team that shares our commitment to patients, to continue to enhance our platform,

and to rapidly advance our precision medicine pipeline of product candidates. The key elements of our strategy are to:

Rapidly advance our lead precision oncology programs, RLY-4008 and RLY-PI3K1047, through clinical development and regulatory approval. We believe our lead precision oncology programs have the potential to treat a wide variety of cancers either as monotherapy or in combination regimens. In September 2020, we initiated a first-in-human clinical trial of RLY-4008. In 2021, we expect to have early safety and efficacy data for RLY-4008 and to be in IND-enabling studies for our RLY-PI3K1047 program. For our wholly-owned programs, we plan to conduct our clinical studies in genetically-defined patient populations. To potentially mitigate development risks, we will leverage learnings from recently approved precision oncology drugs to inform the clinical and regulatory pathways for our lead oncology programs. If we are successful in achieving clinically meaningful anti-tumor activity across solid tumor types, we plan to meet with regulatory authorities to discuss expedited regulatory approval strategies.

Continue to enhance our unique drug-discovery platform. Our Dynamo platform uniquely integrates a broad range of leading-edge experimental and computational technologies and tools, providing us with fundamental insights into the conformational dynamics of target proteins. We are committed to continuously integrating new computational and experimental tools, technologies and capabilities to enhance the power of our Dynamo platform.

Harness the insights and data generated from our platform against intractable targets in oncology and other therapeutic areas. We have built a drug discovery process that leverages our collaboration with D. E. Shaw Research and their access to the Anton 2 supercomputer and our proprietary computational workflows. We are committed to deploying our Dynamo platform against targets in additional therapeutic areas beyond oncology. Our next focus, outside of oncology, is on rare genetic diseases where protein targets are genetically validated, where defects in protein conformational dynamics are abundant, and where we believe our approach is well-suited to identify therapies with the potential to have transformative impact for patients.

Selectively enter into strategic collaborations to maximize the value of our platform and pipeline.We have initiated a Phase 1 clinical trial for RLY-1971 in patients with advanced solid tumors in the first quarter of 2020 and have continued to advance the clinical development of RLY-1971.In December 2020, we entered into the Genentech Agreement, a global collaboration and license agreement with Genentech for the development and commercialization of RLY-1971. Other than our SHP2 program, we retain full development and commercialization rights to our current pipeline of precision medicine programs. We intend to build a fully integrated biopharmaceutical company and independently pursue the development and commercialization of our key product candidates. Given our potential to generate novel product candidates addressing a wide variety of therapeutic indications, we may enter into additional strategic partnerships around certain targets, product candidates, disease areas or geographies. If we believe these collaborations could accelerate the development and commercialization of our product candidates, and allow us to realize additional potential in our product candidates and our platform.

Our Dynamo Platform

The continued and rapid development of new experimental techniques, such as room-temperature crystallography, and computational techniques, such as molecular dynamics and machine learning, is now enabling the deep understanding of protein motion to discover new therapeutic agents. Dynamo was built to capitalize on these recent advances to develop medicines against protein targets with greater specificity and potency. Using our Dynamo platform, we pivot from industry standard SBDD, which is based on static structures and often relies on incomplete protein fragments, to a novel drug-discovery paradigm based on fundamental insights into protein motion, which we term Motion Based Drug Design, or MBDD. We leverage insights from our platform to develop novel, motion-based hypotheses for how to drug target proteins. We can then more rapidly identify and optimize effective lead compounds by integrating powerful experimental and computational tools to sample a much broader range of chemical space than is possible using conventional approaches, which are labor intensive and require significant experimental effort.

In 2016, the confluence of three forces — the proliferation of readily available genomic data, the evolution of experimental techniques, and advances in computational power and speed — led to the founding of Relay Therapeutics. We believe we are uniquely positioned to consolidate these advances and, when combined with our world-class team of experimental and computational experts, integrate these solutions in motion-based drug discovery.

Our platform integrates a broad and tailored array of leading-edge experimental and computational approaches to gain fundamental insights into protein function (Figure 1).

Figure 1: Dynamo drug-discovery platform integrates leading-edge experimental and computational tools.

Key Drug Discovery Steps of our Dynamo Platform

We deploy the power of our Dynamo platform in three key phases of MBDD discovery (Figure 2). We first generate a target modulation hypothesis by developing a detailed mechanistic understanding of the dynamic behavior of the target protein and by identifying pockets where binding of a small molecule can impact protein function. Our platform then aids in the efficient generation of lead compounds through an integrated system of experimental and virtual screens. This enables rapid lead optimization by computationally prioritizing compounds

for experimental evaluation. As each cycle generates new learnings for both our team and our underlying machine learning models, our successful iteration of this process continuously improves our understanding of protein motion which leads to a more effective and efficient drug discovery process.

Figure 2: Dynamo can be deployed across the various stages of drug discovery to provide novel insights and accelerate drug discovery.

Target Modulation Hypothesis

Our first step is to establish a target modulation hypothesis for our protein target of interest.

For each target, the initial goal is to better understand the structure and conformational dynamics of all domains of a protein to generate the target modulation hypothesis. The process typically begins by expressing full-length proteins so we can fully understand the roles of specific domains and accurately capture the differences between the wild-type and mutant forms of the protein (or of different isoforms, etc.). We use a range of leading-edge structural biology techniques (e.g., room temperature X-ray crystallography, Cryo-EM) to visualize these protein conformations in the most physiologically representative context possible. The resulting data allow us to better visualize full-length proteins at atomic resolution. This comprehensive and dynamic visualization enables us to identify potential areas of interest in a protein target that can be exploited in the drug discovery process.

Using a range of protein visualization methods, we can generate a rich experimental understanding of the dynamic conformations of the target protein of interest. We can deploy these experimental data sets in an industry-leading computational platform to generate virtual simulations (molecular dynamics) of the full-length protein moving over long timescales. Long timescale molecular dynamics, or MD, simulations informed by the experimentally derived protein structural data help us better understand how proteins move and change shape over time. Our collaboration with D. E. Shaw Research provides us with access to Anton 2, their proprietary supercomputer that was custom-built for performing molecular dynamics simulations – a technique that calculates the forces between each atom and every other atom in a given system at discrete time points in order to model behavior over time. We use MD simulations to predict the behavior of a given protein system, and with our collaborators we have simulated systems of up to 1 million atoms at time slices of 2.5 x10-15 seconds. The individual time slices are then stitched together to create a high definition movie of the target protein over biologically relevant timescales, typically tens of

microseconds. Other drug discovery approaches may use molecular dynamics, but they are limited to less than 1/100th of the timescale of our simulations. A 10 microsecond simulation of a 1 million atom benchmark protein (satellite tobacco mosaic virus), which requires one day of processing on the Anton 2, would require 271 days on conventional hardware (Nvidia V100) (Figure 3).

Figure 3: The Anton 2 supercomputer enables Relay Therapeutics to simulate the motion of significantly larger biomolecules in far shorter periods of time compared to conventional forms of computation (e.g., GPUs and cloud computing).

After understanding the dynamics of the target protein, we focus on identifying mechanisms to modulate the protein with a small molecule drug. There are multiple ways that a small molecule drug may bind to a target protein to impact its function. Molecules bind to a protein by interacting with amino acids which are often situated in a cavity on the protein’s surface, called a pocket. Most small molecule drugs modulate the function of the target protein by binding to the pocket that directly mediates the protein’s activity, which is called an “active” site. We leverage our platform to identify novel pockets that are not the active site but do impact protein function, so called “allosteric” sites. These binding sites are often part of an allosteric regulatory network that we can elucidate through a combination of computationally derived hypotheses and laboratory experiments on full-length proteins. Our ability to identify novel druggable pockets that have not previously been observed provides new handles for gaining isoform or mutant selectivity.

Our understanding of protein motion and modulation from our Dynamo platform informs the strategy and tools we employ for hit finding and lead generation phases.

Hit Finding and Lead Generation

Once we have identified potential binding pockets and established a target modulation hypothesis, we then transition into hit finding and lead generation, where the goal is to identify a molecule that can serve as the starting point for a new drug.

Our Dynamo platform leverages our motion-based functional understanding of target proteins to enable the design of physiologically relevant activity-based and ligand-centric screens. These experimental measurements of biochemical or biophysical activity are then used to identify molecules to modulate our protein targets. Our Dynamo platform encompasses a variety of screening techniques to identify chemical starting points.

In parallel to our experimental screening efforts, we have made investments in our infrastructure that enables us to use cloud-computing to screen billions of molecules from a virtual compound library in days. The vast number of virtual molecules enables us to sample a much wider diversity of chemical space than would be possible through conventional methods.

Given the powerful hit finding approaches we utilize, we are able to generate a broad diversity of novel small molecules that act via our motion-based target modulation hypothesis and are ready to progress into lead optimization.

Lead Optimization

Once we have identified a chemical starting point and generated a lead compound, optimization is necessary to obtain a molecule that has the desired characteristics. Our Dynamo platform uses advanced computational models in tight integration with our medicinal chemistry, structural biology, enzymology and biophysics capabilities to predict and design the compounds that will achieve the most desirable characteristics, including potency, selectivity, bioavailability, and drug-like properties. Conventional optimization of small molecule lead compounds involves a highly iterative process that includes designing and synthesizing thousands of closely related compounds and experimentally testing them in the lab. This process is time consuming and requires significant experimental effort and expense.

During optimization, we leverage long timescale MD simulations to study binding pocket dynamics and to test analogs of our lead compound to prioritize which ones to synthesize and test experimentally.

Once we have made and tested compounds in the lab, we can compare them to our computational predictions. Over time we can improve our computational predictions using the data that we generate experimentally. We believe that this integration of our long timescale molecular dynamics simulations with experimental data accelerates our lead optimization process.

The Anton 2 supercomputer, that we access through our collaboration with D. E. Shaw Research, makes it possible to run thousands of simulations, which generate vast datasets. To take maximum advantage of this data, we use machine learning algorithms to establish relationships between molecular interactions observed in the simulations and biological activity observed in experiments. In Figure 4, we show how a machine learning model can be trained based on multiple parameters, including ligand motion, protein-ligand interactions and protein conformation, collected during long timescale MD simulations of molecules interacting with our target protein. This model can then be used to make predictions to prioritize the synthesis of new molecules.

Figure 4: Data from long timescale molecular dynamics simulations are used to train machine learning models that can prioritize the next set of molecules to test.

Benefits of Dynamo Platform

Our Dynamo platform was built with the belief that integrating leading-edge computational and experimental approaches would unlock new insight about protein dynamics and ultimately the drug discovery process. We have shown multiple times that we can use this approach to develop novel target modulation hypotheses, generate a broad range of molecular starting points, and rapidly optimize potential drugs. Figure 5 illustrates the timelines for our first two programs relative to conventional drug discovery. In general, it takes three to five or more years to advance from a validated hit to a development candidate, or DC. For our programs, however, we were able to advance from

hit to DC in two years for RLY-1971 and 18 months for RLY-4008. In addition to the advantage of speed, as compared to conventional SBDD, our platform enables us to explore a greater diversity of chemical space, as illustrated by the number of chemical series. This breadth increases the intellectual property landscape that we cover and improves our ability to identify development candidates with optimal drug-like properties.

Figure 5: The Relay Therapeutics Dynamo platform compared to conventional drug discovery approaches.

Our Therapeutic Opportunity

While our Dynamo platform could potentially be applied to a wide range of disease-associated protein targets, we currently focus on precision medicine targets, for which alterations in specific genes are known to cause disease. The genetic diseases we pursue include cancers with clear genetic driver alterations in the tumor genome, as well as monogenic diseases where the causal mutations are present at birth.

Precision Oncology

Our initial focus is in the area of precision oncology where we have seen initial proof of platform in our leading precision oncology pipeline. Over 125 genetic driver alterations across 10 canonical cellular signaling pathways have been identified in 89% of tumors. Targeting these genetic drivers could lead to clinically meaningful responses in patients. However, most of these targets have been intractable to conventional drug discovery approaches or are inadequately drugged by approved therapies. We believe our platform has the potential to address many of these targets by leveraging novel insights into protein dynamics.

Monogenic Diseases

Thousands of monogenic (change in a single gene) diseases exist and affect millions of individuals worldwide. Over 4,000 individual genetic drivers, and their associated protein defects, cause over 7,000 rare monogenic phenotypes. However, since 1996, the U.S. Food and Drug Administration, or FDA, has approved fewer than 70 therapies to specifically treat these conditions, presenting a vast unmet therapeutic need. We believe our Dynamo platform has the potential to address many of these targets.

Other Precision Medicines Opportunities

The decreasing cost and increasing resolution of genomic data have identified hundreds of additional actionable genetic targets beyond precision oncology and monogenic disease. These include genetically-defined subpopulations

of more common diseases in neurology, immunology and other therapeutic areas. We believe that there are multiple genetically-validated targets in these disease areas that are unaddressed by approved therapies, representing an area of significant unmet need. We believe our Dynamo platform has the potential to address many of these targets.

Our focus on addressing the genetic drivers of disease, also referred to as genetically validated targets, confers several advantages, including:

Clear causal link to disease: Genetic diseases offer an unambiguous causal link between the mutational alteration in a specific gene, disease biology, and a patient’s symptoms, such that the translational medicine hypothesis is well-validated at the beginning of a drug discovery program.

Precision medicine opportunity: Because of the strong link between specific genetic alterations and disease symptoms, it is possible to precisely target therapy to genetically identifiable patients who are most likely to respond favorably to a precision medicine.

Increased translational success: We believe that the ability to precisely target therapy to patients who are most likely to respond favorably to treatment will, in turn, increase the likelihood of successfully translating a specific pharmacological mechanism into clinical benefit.

While we have initially focused our efforts on small molecule therapies, our Dynamo platform could also be readily deployed towards the discovery of other therapy types, such as large molecules including peptide or protein therapeutics.

Our Product Pipeline and Programs

We have deployed our Dynamo platform to initially focus on the area of precision oncology. To date, we have generated several promising precision oncology, orally available, small molecule product candidates that address previously intractable oncogenic targets. Our lead programs are targeting a range of driver alterations to treat various cancers that we believe can have a greater probability of translational success because they are genetically or clinically validated. The targets associated with all of our current programs are Category 1 Targets under our DESRES Agreement. See “—Our Collaborations—License Agreements and Strategic Collaborations —Collaboration and License Agreement with D. E. Shaw Research, LLC.” In addition, we are also advancing several early programs focused on other precision oncology and rare genetic disease targets. In December 2020, we entered into the Genentech Agreement, a global collaboration and license agreement with Genentech for the development and commercialization of RLY-1971. See “—Our Collaborations—License Agreements and Strategic Collaborations—Collaboration and License Agreement with Genentech.” Other than our SHP2 program, we retain full development and commercialization rights to our current pipeline of precision medicine programs.

The following table summarizes our current portfolio of product candidates and programs.

RLY-4008, a selective inhibitor of FGFR2

Overview

RLY-4008 is designed to be an oral, small molecule, selective inhibitor of fibroblast growth factor receptor 2, or FGFR2, a receptor tyrosine kinase that is frequently altered in cancer. FGFR2 is one of four members of the FGFR family, a set of closely related proteins with highly similar protein sequences and properties. RLY-4008 minimally inhibits targets other than FGFR2 and demonstrates FGFR2-dependent cell-killing in cancer cell lines. We initiated a first-in-human clinical trial of RLY-4008 enriched for patients with advanced solid tumors having oncogenic FGFR2 alterations in September 2020. We believe FGFR2-mediated cancers affect approximately 8,000 late-line patients annually in the United States, of which fusions represent approximately 2,700, amplifications approximately 1,600, and mutations approximately 3,800. In the future, if RLY-4008 advances to earlier lines of treatment, we believe it could potentially address approximately 20,000 patients annually in the United States across the different alterations.

Role of FGFR in cellular proliferation and differentiation

Each of the FGFRs has an important role in normal physiology and the inhibition of FGFR2 is a well-validated pathway in disrupting cancer proliferation and growth. Two non-selective FGFR inhibitors have been approved (erdafitinib and pemigatinib) and several are in clinical development. However, these inhibitors as a class cause several dose-limiting, FGFR2-unrelated toxicities in patients leading to dose reductions and altered dosing schedules. One of the most common dose limiting toxicities of these agents is hyperphosphatemia (buildup of excess phosphate in the bloodstream), which causes soft tissue mineralization and requires active management. Hyperphosphatemia has been shown to be driven by inhibition of another member of the FGFR family known as FGFR1 (Figure 15).

Figure 15: RLY-4008 is a selective inhibitor of FGFR2. FGFR1 is required for phosphate resorption in the kidney. Inhibition of FGFR1 by non-selective FGFR inhibitors results in increased serum phosphate and toxicity. This results in decreased efficacy by requiring dose reductions.

We believe that the toxicity attributable to inhibition of other FGFR family members, and other closely related kinases, limits the ability of the non-selective FGFR inhibitors to achieve optimal and durable inhibition of FGFR2, limiting the efficacy of these agents in patients with FGFR2-altered tumors. In addition to the lack of selectivity, these inhibitors are unable to overcome on-target resistance, which has been observed in patients treated with non-selective FGFR inhibitors. Our belief is that a selective inhibitor of FGFR2 that retains activity against resistance mutations will enable improved clinical efficacy.

Limitations of current FGFR inhibitors

Non-selective FGFR inhibitors produced by other companies have demonstrated clinical proof-of-concept in patients with intrahepatic cholangiocarcinoma, or ICC, bearing FGFR2 gene fusions. These gene fusions result in a constitutively active FGFR2, which promotes oncogenic transformation. Genetic alterations in FGFR2, including gene fusions, amplifications, and point mutations, are also found in other solid tumor indications.

Patients with genetic alterations in FGFR2, primarily gene fusions in ICC, have been treated with FGFR inhibitors in investigational clinical trials. To date, these trials provide support for the critical role of FGFR2 for tumor survival with a response rate of up to 36% (Figure 16). A key limiting factor for existing FGFR therapies is that, as a class, they are associated with a dose-limiting side effect, hyperphosphatemia, which has been shown to be caused by FGFR1 inhibition.

Figure 16: Hyperphosphatemia is a dose-limiting adverse event associated with non-selective FGFR inhibitors.

Our solution, RLY-4008

RLY-4008 is an oral, small molecule inhibitor of FGFR2 designed to inhibit FGFR2 with high potency while minimizing inhibition of other FGFR family members. In our initial assessment of the challenge of obtaining a highly selective inhibitor of FGFR2, we determined that there is a high degree of structural similarity between FGFR1 and FGFR2 when comparing static X-ray crystal structures. This similarity precluded the development of a structure-based selectivity hypothesis using conventional approaches.

We therefore set out to identify motion-based differences between FGFR2 and other FGFR family members by applying our expertise in computational modeling and experimental structural analyses. We discovered that there were segments of FGFR2 which displayed differential dynamics compared to the corresponding segments of FGFR1 (Figure 17). We predicted these dynamic differences could be exploited to achieve selective inhibition of FGFR2.

Figure 17: Using MD simulations, we predicted that a segment in FGFR1 was more dynamic than FGFR2, as represented by the schematic below where the segment opens “Up” more frequently in FGFR1 compared to FGFR2.

We embarked on a process using computational methods such as long timescale molecular dynamics simulations, virtual docking and specialized experimental techniques to design, select, synthesize, and evaluate inhibitors. Our discovery process culminated with the selection of RLY-4008 as a product candidate based on its ability to meet our predetermined criteria for potency, selectivity and activity in animal models.

As described below, we have conducted a number of head-to-head preclinical experiments utilizing cellular in vitro assays and mouse models using equivalent methods on all compounds tested. We compared RLY-4008 across a number of parameters against two approved molecules as well as three molecules in clinical development. We believe that the results of these preclinical experiments have demonstrated that RLY-4008 could potentially be a differentiated molecule warranting testing in clinical studies. We initiated a first-in-human clinical trial of RLY-4008 enriched for patients with advanced solid tumors having oncogenic FGFR2 alterations in September 2020. Given RLY-4008’s strong preclinical activity against both primary oncogenic alterations and acquired pan-FGFR inhibitor resistance mutations, the trial will include patients that are naïve to pan-FGFR inhibitors, as well as those that have progressed on pan-FGFR inhibitors.Ultimately, extensive clinical studies of RLY-4008 will be required to determine if the differentiation we observed in the preclinical studies described below translates into clinical benefit for patients. The clinical data that we expect to generate in any such clinical studies will constitute the bulk of the data needed to support an application for marketing approval of RLY-4008. Unless we conduct head-to-head studies of RLY-4008 against other molecules as part of our future clinical trials, we would not expect to rely upon RLY-4008’s potential differentiation from any other molecules in connection with submissions to the FDA or other regulatory agencies, as applicable, for approval or otherwise.

In addition, in our head-to head preclinical experiments utilizing cellular in vitro assays and mouse models, we selected dose levels for the other molecules to match human exposure of the approved dose or the dose being used in its clinical studies, as applicable, and we used the proportional dose levels of RLY-4008 that we believe are comparable to what we expect to utilize in our future clinical studies. The differences in these dose levels may have had an impact on the differentiation in the preclinical results we observed.

We demonstrated in enzymatic and cellular assays that RLY-4008 was over 200-fold more potent at inhibiting FGFR2 compared to FGFR1 (Figure 18). In addition to selectivity over FGFR1, RLY-4008 is also selective over the other members of the FGFR family, FGFR3 (>80-fold) and FGFR4 (>4,000 fold) in biochemical assays.

Figure 18: RLY-4008 is selective for FGFR2 over FGFR1.

The selectivity of RLY-4008 for FGFR2 over FGFR1 was determined by comparing the potency (IC50) of RLY-4008 and other clinical non-selective FGFR inhibitors in biochemical assays using Caliper technology (PerkinElmer). Human FGFR1 and FGFR2 (Carna Biosciences) were incubated with a peptide substrate (PerkinElmer) in the presence of varying concentrations of the indicated inhibitor for 30 minutes. Reactions were carried out in the presence of 100 mM ATP and 10 mM MgCl2 for 90 minutes. Non-selective FGFR inhibitors were obtained from vendors that provide compounds based on chemical structures published in the patent literature (Cayman Chemical, MedChemExpress, Selleckchem). Fold change in potency was calculated using the average IC50 obtained for each inhibitor in three independent experiments. RLY-4008 showed greater than 200-fold selectivity for FGFR2 over FGFR1.

RLY-4008 has minimal inhibition of targets other than FGFR2 and demonstrates FGFR2-dependent cell-killing in cancer cell lines. It has bioavailability suitable for oral dosing, is metabolically stable, and has demonstrated good

pharmacokinetics in preclinical in vivo models. Human pharmacokinetic projections are consistent with once or twice daily oral dosing. RLY-4008 is predicted to have low risk of drug-drug interactions based on weak inhibition of drug metabolizing enzymes. It is readily synthesized in bulk, can be formulated for oral delivery, and exposures at the highest non-severely toxic dose were several fold in excess of the predicted human efficacious exposures.

In a patient-derived xenograft, or PDX, mouse model of ICC harboring a FGFR2 fusion, treatment with RLY-4008 led to tumor regression at doses as low as 10 mg/kg delivered twice a day (Figure 19). Non-selective inhibitors, pemigatinib and futibatinib, also resulted in tumor volume reductions in this model when dosed at levels selected to match their human exposure in clinical studies.

To preclinically validate our effort to engineer selectivity for FGFR2 as a means of reducing the risk of hyperphosphatemia, we examined the effect of RLY-4008 in an industry standard rat model of hyperphosphatemia. No evidence of hyperphosphatemia was seen with doses of RLY-4008 that resulted in exposures leading to tumor regression in our FGFR2 gene fusion ICC PDX mouse model (Figure 19). By contrast, when dosed at levels selected to match human exposure in clinical studies, futibatinib led to increased hyperphosphatemia. Additionally, in 28-day GLP toxicology studies in rats and dogs, neither hyperphosphatemia nor tissue mineralization were observed with RLY-4008 at exposures in the animal corresponding to the predicted human efficacious exposures.

Figure 19: RLY-4008 leads to tumor regression in an FGFR2 fusion positive ICC PDX model and does not cause hyperphosphatemia.

(a) Anti-tumor activity of RLY-4008 dosed twice daily (BID) by oral administration in an FGFR2 fusion-positive ICC PDX model. RLY-4008 induced dose-dependent regression when administered at 10 or 30 mpk BID. TAS-120 at 2 mpk TID (red) and pemigatinib at 0.5 mpk BID (pink) were dosed at levels selected to match their clinical exposures. Data points indicate mean tumor volume (n=6 per group) and error bars represent standard error of the mean. All treatment groups are statistically significant when compared to vehicle with p<0.001 as determined by two-sided t-test.

(b) Serum phosphate measurements in rats dosed twice daily with RLY-4008 (blue) or TAS-120 (red) by oral administration. Doses of RLY-4008 (100 and 200 mpk BID) resulting in exposures leading to tumor regression in our FGFR2 gene fusion ICC PDX model do not cause significant hyperphosphatemia. Doses of TAS-120 (3 mpk BID and 10 mpk QD) selected to match human exposures in clinical studies cause significant hyperphosphatemia. Data indicate the mean serum phosphate level (n=5 per group), and error bars represent standard deviation. TAS-120 treatment groups are statistically significant when compared to vehicle with p<0.01 as determined by one-way ANOVA.

Additionally, RLY-4008 was able to achieve in vivo efficacy in mouse models of FGFR2-fusion gastric cancer, FGFR2-amplified gastric cancer, and FGFR2-mutant endometrial cancer. Treatment with RLY-4008 led to tumor regression at 3 mg/kg delivered twice a day in a FGFR2 gene fusion gastric cancer model, and at 10 mg/kg delivered twice a day in a FGFR2-amplified gastric cancer or FGFR2-mutant endometrial cancer models (Figure 22). All of these doses result in exposures that do not cause hyperphosphatemia in an industry standard rat model. Importantly, RLY-4008 achieved complete regression in an FGFR2-mutant endometrial cancer model (AN3CA) harboring the N550K mutation that reduced the potency of pemigatinib by 185-fold.

Figure 22: RLY-4008 leads to tumor regression in an FGFR2-fusion gastric cancer PDX, the FGFR2-amplified gastric cancer SNU16 xenograft model, and the FGFR2 N550K-mutant endometrial cancer AN3CA xenograft model.

Anti-tumor activity of RLY-4008 dosed twice daily (BID) by oral administration in (a) an FGFR2 fusion gastric cancer PDX model, (b) the FGFR2-amplified SNU16 gastric cancer xenograft model, and (c) the FGFR2 K310R; N550K mutant AN3CA endometrial cancer xenograft model. Data points indicate mean tumor volume and error bars represent standard error of the mean. Statistical analyses were performed using one-way ANOVA. (a) n=8 per group; treatment groups are statistically significant when compared to vehicle with p<0.001. (b) n=7 per group; treatment groups are statistically significant when compared to vehicle with p=0.001 for 10 mpk BID group and p<0.001 for 30 mpk BID group. (c) n=8 per group; two high-dose groups are statistically significant when compared to vehicle with p=0.003 for 30 mpk BID group and p=0.005 for 10 mpk BID group. p=0.627 for 3 mpk BID group.

Another predicted advantage of RLY-4008 concerns resistance mutations. These new mutations in FGFR2 arise during treatment, reducing the potency of non-selective FGFR inhibitors and making tumors resistant to treatment. In preclinical experiments, we have shown that RLY-4008 retains activity against a broad panel of mutations known to be associated with resistance to non-selective FGFR inhibitors (Figure 20).

Figure 20: RLY-4008 retains potency against common FGFR2 resistance mutations.

Heatmap showing fold change in potency (IC50) on FGFR2 mutations compared to FGFR2 WT. Human FGFR2 cDNA (NCBI Reference Sequence: NM_022970.3) was cloned into pLenti-P2A-Puro vector, site-directed mutagenesis was performed for the indicated mutations, and plasmid purification was conducted at GenScript. Lentivirus for each vector was prepared using Lenti-vpak Lentiviral Packaging Kit, and FGFR2 WT or FGFR2 mutants were expressed in HEK-293 cells via lentiviral transduction. Cells were incubated with various concentrations of the indicated inhibitors for 2 hours and potency of FGFR2 was determined using a pFGFR2 (Tyr 653/654) HTRF assay (Cisbio) per the manufacturer’s protocol. Colors indicate the fold loss in potency for the mutant FGFR2 vs WT. Gatekeeper mutations block access to the binding site of non-selective inhibitors. Molecular brake mutations disrupt an autoinhibitory conformation of FGFR2, resulting in kinase activation. Other mutations listed have various reported mechanisms of kinase activation.

In the studies published to date describing resistance, multiple FGFR2 resistance mutations have been reported, with mutations at position V565 and N550 being most common. Mutations sterically block access to the binding site of non-selective FGFR inhibitors and/or disrupt an auto-inhibitory conformation of FGFR2. Among mutations, V565F and N550K are two of the most prevalent. To further evaluate the activity of RLY-4008 against FGFR2 resistance mutations, the in vivo activity of RLY-4008 was compared to five different pan-FGFR inhibitors including pemigatinib, infigratinib, erdafitinib, zoligratinib, and futibatinib in an ICC FGFR2 fusion cancer PDX model with the V565F mutation and to pemigatinib in the AN3CA endometrial cancer CDX model with the N550K mutation. In both models, RLY-4008 was able to induce regression at doses resulting in exposures that do not cause hyperphosphatemia in an industry standard rat model. By contrast, when dosed at levels selected to approximate human exposure in clinical studies, the pan-FGFR inhibitors showed little to no anti-tumor activity in both models (Figure 23).

Figure 23: RLY-4008 induces tumor regression in FGFR2 V565F and N550K-mutant xenograft models.

Anti-tumor activity of RLY-4008 and pan-FGFR inhibitors in xenograft models harboring common on-target resistance mutations. (a) RLY-4008 and pan-FGFR inhibitors were dosed once daily (QD) by oral administration in an ICC FGFR2 fusion cancer PDX model with the V565F mutation. RLY-4008 induced regression whereas all pan

FGFR inhibitors grew in the presence of drug.(b) RLY-4008 and pemigatinib dosed twice daily (BID) by oral administration in the FGFR2 K310R;N550K mutant AN3CA endometrial cancer xenograft model. RLY-4008 induced regression whereas pemigatinib was inactive. Data points indicate mean tumor volume and error bars represent standard error of the mean. RLY-4008 cohort is statistically significant when compared to vehicle or pan-FGFR inhibitors with p<0.001 as determined by one-way ANOVA.

Although RLY-4008 retains activity preclinically against common FGFR2 resistance mutations, tumors may develop bypass resistance by shifting growth factor signaling to an alternate receptor, rendering them less sensitive to the targeted therapy. SHP2, a protein tyrosine phosphatase, regulates the activity of multiple RTKs, and may be an effective way to overcome bypass resistance to RLY-4008. To demonstrate the potential for RLY-1971 as a combination partner for RLY-4008, we tested a population of four patient-derived FGFR2-fusion positive ICC cell lines. These cells were derived from patients that initially responded to non-selective FGFR inhibitors, but then acquired bypass resistance to FGFR inhibition during their treatment. While these cell lines were resistant to treatment with our FGFR2 inhibitor RLY-4008, all resistant cells were sensitive to RLY-1971 with IC50s of less than 100 nM (Figure 24). Given the role of SHP2 in mediating bypass resistance to multiple targeted therapies, we intend to investigate the clinical potential of the combination of RLY-1971 with RLY-4008.

Figure 24: RLY-1971 overcomes bypass resistance to FGFR2 inhibition in patient-derived FGFR2 fusion positive ICC cell lines.

Anti-proliferative effect of RLY-1971 in patient-derived FGFR2 fusion positive ICC cells tested in a 2D proliferation assay. These cell models were derived from patients that initially responded to non-selective FGFR inhibitors, but then progressed during their treatment (the specific FGFR2 fusion present in the cells is indicated). These cells are resistant to treatment with our FGFR2 inhibitor RLY-4008 (blue lines), whereas RLY-1971 demonstrates anti-

proliferative and cytotoxic activity (red lines), with IC50s as follows: (a) 13 nM, (b) 20 nM, (c) 39 nM, and (d) 91 nM. The dotted line at 0 indicates complete growth suppression, with values below 0 indicating cytotoxicity.

Our clinical development plan

The RLY-4008 clinical development plan seeks to leverage the unique potential for enhanced tolerability and broad FGFR2 mutational coverage to rapidly generate proof-of-concept in molecularly defined patient subsets.

We initiated a first-in-human clinical trial of RLY-4008 enriched for patients with advanced solid tumors having oncogenic FGFR2 alterations in September 2020. The primary objectives are to determine the maximum tolerated dose (MTD)/recommended phase 2 dose (RP2D) and to define the overall safety profile of RLY-4008. Secondary objectives are to assess the pharmacokinetics, pharmacodynamics and to explore anti-tumor activity of RLY-4008. Patients will initially receive RLY-4008 administered orally, twice daily.

The first trial will employ a 2-part dose escalation/dose-expansion design. Given RLY-4008’s strong preclinical activity against both primary oncogenic alterations and acquired pan-FGFR inhibitor resistance mutations, the trial will include patients that are naïve to pan-FGFR inhibitors, as well as those that have progressed on pan-FGFR inhibitors. Observation of significant clinical activity in one or more patient populations in this exploratory first-in-human trial would support further trials to confirm the risk-benefit profile of RLY-4008 in patients with oncogenic FGFR2 alterations. These trials may include continued evaluation of RLY-4008 as a monotherapy in single arm trials in patient populations without an established standard-of-care therapy available, which could be used to support filings for marketing authorization for RLY-4008. The development program for RLY-4008 may also include randomized trials of RLY-4008 compared to a relevant standard-of-care therapy.

We anticipate giving an initial clinical update on this trial in the second half of 2021.

Figure 25: First-in-human clinical trial for RLY-4008

Development of RLY-4008 will require identification of appropriate patients for treatment with FGFR2 alterations using molecular diagnostic tests. In early phase clinical trials, patients will be identified using local testing performed at clinical trial sites, with retrospective centralized testing to confirm the tumor genetic status. In later phase trials, we will likely collaborate with a diagnostic partner to identify patients for clinical trial enrollment using an analytically validated investigational molecular diagnostic. The tumor genetic contexts that we are considering for development of RLY-4008 (FGFR2 fusions, amplifications and mutations) can currently be detected using FDA-approved next generation sequencing based panel diagnostics (e.g. Foundation One, Guardant 360).

Mutant-PI3Kα Inhibitor Programs

Overview

RLY-PI3K1047 is the lead compound in our franchise of programs targeting cancer-associated mutant variants of phosphoinositide 3-kinase alpha, or PI3Kα. RLY-PI3K1047 is a small molecule inhibitor of PI3Kα that we designed to specifically target PI3Kα H1047X mutant via a previously undescribed allosteric mechanism. Oral dosing of RLY-PI3K1047 resulted in tumor growth inhibition in a mouse xenograft model of PI3Kα H1047R mutant carcinoma. We expect to begin IND-enabling studies for a differentiated PI3Kα H1047X mutant-selective inhibitor in 2021. We believe PI3Kα H1047X mutant cancers affect approximately 10,000 late-line patients annually in the United States. In the future, if RLY-PI3K1047 advances to earlier lines of treatment, it could potentially address approximately 50,000 patients annually in the United States.

Figure 26: PI3Kα addressable patient populations

1) Estimated frequency percentages are based on counts of known/likely functional alterations in the Foundation Medicine Insights database. 2) Based on projected cancer deaths in all solid tumors from the National Cancer Society’s SEER database as a proxy for late-line cancer patient incidence. 3) These data are based on projections from the National Cancer Society’s SEER program for estimated new cases of advanced solid tumors.

Role of PI3Kα in cellular proliferation and differentiation

Mutations at amino acid H1047 of PI3Kα are among the most common kinase mutations in cancer and are believed to be a primary driver of carcinogenesis. There are no approved therapies that selectively target mutant versions of PI3Kα. Inhibitors that are not mutant-selective are associated with dose-limiting toxicities resulting in frequent discontinuations that restrict their therapeutic potential. Additionally, these inhibitors also can inhibit other isoforms of PI3K, including PI3Kδ, which can further result in toxicity. Our belief is that selectively targeting mutant PI3Kα could result in improved target inhibition and increased clinical efficacy (Figure 27).

Figure 27: RLY-PI3K1047 is a selective inhibitor of H1047X mutant PI3Kα. WT PI3Kα plays a critical role in normal cellular signaling and function, including glucose uptake and insulin regulation. Inhibition of WT PI3Kα by non-mutant selective PI3K inhibitors results in hyperglycemia, hyperinsulinemia and other toxicities. This results in decreased efficacy by requiring dose reductions.

Leveraging our structural biology capabilities, we solved what we believe to be the first full-length structure of PI3Kα using cyrogenic electron microscopy (Cryo-EM) and utilized a range of experimental techniques to understand both H1047R mutant and wild-type conformations. We used this rich experimental data set to power molecular dynamics simulations of H1047R mutant PI3Kα to identify a series of dynamic structural changes caused by the mutation, which were not elucidated by prior structural studies of either H1047R mutant or wild-type PI3Kα. The lead compound in this program, which we refer to as RLY-PI3K1047, was designed to exploit these dynamic differences and bind to a novel allosteric site to achieve heightened mutant selectivity. We intend to initiate IND-enabling studies for our first PI3Kα mutant selective inhibitor, which is focused on H1047X, in 2021.

PI3Kα mutations drive the development of cancer

PI3Kα is the central regulator of a cellular signaling pathway that has been linked to a diverse group of cellular functions related to cancer including cell growth, proliferation and survival. Data collected as a part of Foundation Medicine Insights and other data sources identifies PI3Kα as the most frequently mutated kinase in cancer. (Figure 28).

Figure 28: PI3Kα is a common mutation in cancer.

Approximately 80% of the mutations in PI3Kα cluster at three amino acids, or locations. These are E542 and E545 in the helical domain, and H1047 in the kinase domain. The most common mutation at amino acid H1047 is H1047R, but other H1047 mutations (such as H1047L, H1047Y, and others) are also observed across cancers. The abbreviation “H1047X” is used to refer to any H1047 mutation. Similarly, the most common mutations in the helical domain are E542K and E545K, but other mutations (such as E542Q, E545A, and others) are also observed across cancers. The abbreviation “E542/E545X” is used to refer to any helical domain mutation. The H1047R mutation has been shown to induce extensive and diverse cellular changes in pre-clinical models of breast cancer, demonstrating how a single mutation at amino acid H1047 can have large consequences and induce a cancer phenotype. The E5452K and E545K mutations have also been shown to increase PI3Kα activity, promote cell growth and invasion in vitro, and induce tumorigenesis in vivo. While H1047X and E542/E545X mutations have been shown to result in aberrant PI3Kα activity, they do so through distinct biological mechanisms.

Limitations of current PI3Kα inhibitors

Given the large number of patients with PI3Kα mutations, several small-molecule inhibitors of PI3Kα are in development for oncology indications. However, these inhibitors have to our knowledge been largely ineffective when used as monotherapy in cancer. All of these inhibitors target the catalytic (orthosteric) site of PI3Kα. One challenge faced by these inhibitors has been drug intolerance, especially at the high doses routinely used in cancer trials. Alpelisib, marketed as Piqray® by Novartis, is the only FDA-approved inhibitor for cancers with mutated PI3Kα. However, alpelisib is not a selective inhibitor for mutant forms of PI3Kα; it is a potent inhibitor of both the wild-type form of PI3Kα as well as the mutant form. Nonetheless, alpelisib is approved to be used in combination with fulvestrant, an estrogen receptor degrader, in PI3Kα-mutated breast cancer. When used in combination with fulvestrant, alpelisib was associated with significant adverse events, including severe hypersensitivity, diarrhea and severe pneumonitis. Hyperglycemia was reported in 64% of patients and over 36% of patients experienced Grade 3 or Grade 4 hyperglycemia. To manage hyperglycemia, insulin along with other anti-diabetic medication was used in 87% of patients. Gastrointestinal toxicity was reported in 93% of patients, with 9% experiencing Grade 3 gastrointestinal toxicity. Additionally, 36% of patients experienced rash, with 10% experiencing Grade 3 rash. The combination of these adverse events resulted in 64% of patients requiring dose reductions and 25% of patients discontinuing treatment. Despite 11 month progression-free survival (PFS) in the SOLAR-1 Phase 3 trial of alpelisib, the median duration of dosing in the alpelisib arm was 5.5 months, indicating the majority of patients

discontinued dosing prior to disease progression. The observed hyperglycemia is believed to be caused by inhibition of wild-type PI3Kα and therefore is considered an on-target toxicity for alpelisib. In addition to causing dose-limiting toxicity, systemic glucose-insulin feedback caused by inhibiting wild-type PI3K results in elevated insulin that can activate PI3K signaling and subsequently limit the efficacy of PI3K inhibitors. While these factors limit the clinical utility of alpelisib, these data nonetheless establish mutant PI3Kα as a clinically validated target in breast cancer. Because these toxicities result in suboptimal doses and dosing schedules that result in incomplete PI3Kα inhibition, we believe that a H1047X or E542/E545K mutant selective inhibitor will enable improved target inhibition, and therefore improved clinical efficacy. Additionally, overcoming hyperinsulinemia and hyperglycemia could increase efficacy by preventing insulin feedback that activates PI3K signaling.

Our solution, mutant selective inhibition of PI3Kα

Given the existence of mutations in PI3Kα with different biological mechanisms underlying aberrant activity, we believe there are multiple opportunities to develop distinct mutant selective inhibitors of PI3Kα. Addressing the challenge of mutant selectivity required us to express and then solve the structure of the full-length PI3Kα protein. This structure, which to our knowledge had previously not been solved, represented a technical challenge because PI3Kα is a membrane-bound protein. This type of protein is typically difficult both to purify in large quantities and to crystallize. Nonetheless, we were able to obtain the structure of full-length PI3Kα using Cryo-EM. The three-dimensional structure of PI3Kα was determined by collecting data from two-dimensional electron microscopic projections of thin layers of protein. The resulting three-dimensional protein structure provided us with fundamental insights into the mechanism of activation of PI3Kα and the impact of mutations on its function. Through the integration of these structural insights with a combination of experimental and computational techniques, our aim is to develop a franchise of mutant selective PI3Kα inhibitors. The first lead molecule derived from these efforts, which is focused on H1047X, is described below.

Current lead molecules for PI3K-H1047X Mutations, RTX-1 and RTX-2

RLY-PI3K1047 encompasses our lead small molecule inhibitors of PI3Kα, RTX-1 and RTX-2, that we designed to specifically target PI3Kα H1047X mutant via a previously undescribed allosteric mechanism.As described above, adverse events such as hyperglycemia are common among PI3K inhibitors that have been tested in the clinic, leading us to focus on identifying an inhibitor that bound to a novel site on PI3Kα. Our intent was to obtain a molecule that could selectively bind to the mutant form of PI3Kα.

Structural analyses of PI3Kα showed that mutations at amino acid H1047 cause structural alterations that are located away from the catalytic site, the place where other PI3K inhibitors bind. We then performed long timescale molecular dynamics simulations of wild-type and H1047R mutant PI3Kα to identify a series of dynamic structural changes caused by the mutation that are not present in the wild-type protein.

Utilizing this structural information, we designed inhibitors to target a novel allosteric binding site on the PI3Kα H1047R mutant protein that our computational and experimental approaches exposed. This process led to the discovery of RTX-1 and RTX-2, which are approximately 5-10-fold selective for the H1047R mutant form of PI3Kα compared to the wild-type protein in biochemical assays (Figure 29). In contrast, alpelisib and GDC-0077 (an orthosteric PI3Kα inhibitor currently in development) biochemically inhibited the mutant and wild-type proteins with approximately equivalent potency. In addition, we found that RTX-1 and RTX-2 are selective for PI3Kα over other PI3K isoforms, including PI3Kß and PI3Kδ, showing no measurable inhibition. In contrast, alpelisib and GDC-0077 inhibited the PI3Kδ isoform with IC50 < 1μM. Given toxicities associated with inhibitors that target PI3K isoforms other than PI3Kα and GDC-0077, including gastrointestinal side effects and transaminitis, we believe that these molecules provide a dual advantage of isoform and mutant selectivity, which could result in increased clinical efficacy compared alpelisib or other orthosteric PI3Kα inhibitors.

Figure 29: Compared to alpelisib and GDC-0077, RTX-1 and RTX-2aremore selective for the PI3Kα mutant (H1047R) compared to wild-type (a) and more selective for the PI3Kα isoform compared to other PI3K isoforms PI3Kß and PI3Kδ (b).

Biochemical potency for RLY-PI3K1047 compared to alpelisib and GDC-0077. IC50 values are shown for inhibition of the PI3Kα mutant (H1047R) compared to wild-type (a) and for PI3Kα compared to other PI3K isoforms (b). Phosphotransfer activity (PtdIns(3,4,5)P3 production in liposomes using diC8-PtdIns(4,5)P2 as a substrate in the presence of 100uM ATP and titrated compounds after a 120min incubation) was measured by ADP-Glo. All samples were run in duplicate and data represent the mean.

This increased biochemical potency for PI3Kα H1047R mutant protein translates into an increased potency in cellular pharmacodynamic assays. RTX-1 was approximately 10-fold more potent for inhibition of phosphorylated AKT (pAKT), a key substrate of PI3Kα, in transformed breast epithelial cells expressing PI3Kα H1047R compared to the same cells expressing wild-type PI3Kα. RTX-2, an example of another lead compound generated in this program, also showed approximately 10-fold increased potency for inhibition of pAKT in transformed breast epithelial cells expressing PI3Kα H1047R (Figure 30). In contrast, alpelisib and GDC-0077 (an orthosteric PI3Kα inhibitor currently in development) showed approximately equal potencies in cells expressing either the mutant or wild-type forms.

Figure 30: Compared to other clinical PI3Kα inhibitors (alpelisib and GDC-0077), Relay compounds more potently inhibits pAKT in cells expressing H1047R mutant PI3Kα compared to cells expressing wild-type PI3Kα.

Inhibition of pAKT by Relay compounds RLY-PI3K1047 and RTX-2 in a pharmacodynamic assay. MCF10A immortalized breast epithelial cells endogenously expressing wild-type PI3Kα(a) or engineered to express the PI3Kα H1047R mutation (b) were treated with alpelisib, GDC-0077 or Relay compounds. After 2 hours cell lysates were collected and the impact on pAKT levels was assessed using an HTRF assay (three fold dilution dose response was run in duplicate, data represented as mean +/- standard error of the mean). Half maximal effective concentrations (EC50) from the dose response curves (a, b) are plotted in (c).

The selectivity of RTX-1 was then evaluated in vivo. Oral dosing of RTX-1 resulted in tumor growth inhibition in a mouse xenograft model of PI3Kα H1047R carcinoma at doses of 100 mg/kg delivered once or twice daily or 50 mg/kg delivered twice daily. (Figure 31). An important validation of our efforts to avoid the dose-limiting toxicities associated with other PI3K inhibitors is the effect of RTX-1 on hyperinsulinemia. As discussed above, hyperinsulinemia and hyperglycemia can lead to decreased efficacy of PI3K inhibitors. In a study evaluating the effects of alpelisib or RTX-1 treatment on insulin levels, RTX-1 led to minimal changes in serum insulin (p=0.116 compared to vehicle by 2-way ANOVA) when administered orally at all doses tested for the duration of the study. In contrast, alpelisib treatment resulted in increases in serum insulin (p<0.0001 compared to vehicle by 2-way ANOVA). Additionally, in an oral glucose tolerance test (OGTT) assessing insulin response after dosing of compounds, alpelisib treatment at 50 mg/kg once daily lead to larger increases in serum insulin compared to all doses of RLY-PI3K1047 tested (p<0.0001 by 2-way ANOVA).

Figure 31: RTX-1 inhibits tumor growth in vivo with minimal increases in serum insulin levels.

Anti-tumor activity and impact on serum insulin levels in response to treatment with RTX-1.

(a) The CAL33 xenograft model was dosed once or twice daily (12 hour interval) with RTX-1 by oral administration or alpelisib once daily by oral administration, and tumor growth was evaluated. Data represent mean tumor volume over time, and error bars represent standard error of the mean. (n=8 per group).

(b) Insulin levels in serum were measured one hour before and one hour after drug administration in non-tumor bearing animals throughout an 8 day dosing period (measurements were taken specifically on day 1, 3, 5 and 8, n=8 per group). Data presented as mean +/- standard error of the mean.

While RTX-1 is one lead molecule generated in this franchise, we are continuing lead optimization to identify mutant selective inhibitors of PI3Kα meeting our criteria to enter IND-enabling studies.

Our clinical development plan

We expect to begin IND-enabling studies for a differentiated PI3Kα H1047X mutant-selective inhibitor in 2021. With this profile, we will look to advance a precision medicine program that quickly establishes safety, tolerability, and preliminary efficacy, in patients with advanced solid tumors with H1047X mutations. Upon completion of dose escalation, the mutant PI3Kα inhibitor will be tested as a monotherapy in advanced cancer patients with PI3Kα H1047X mutations in a tumor-agnostic study. We will also pursue disease-specific development paths including combination with endocrine therapy +/- CDK4/6 inhibitors in hormone-receptor positive breast cancer and a PI3Kα H1047X mutation.

RLY-1971, an inhibitor of SHP2

Overview

RLY-1971 is designed to be an oral, small molecule inhibitor of the protein tyrosine phosphatase SHP2 that binds and stabilizes SHP2 in its inactive conformation. SHP2 promotes cancer cell survival and growth through the RAS pathway by transducing signals downstream from RTKs. Additionally, activating SHP2 mutations result in enhanced signaling in the absence of ligand stimulation and has been identified as oncogenic drivers in a range of tumors. As a critical signaling node and regulator, SHP2 drives cancer cell proliferation and plays a key role in the way cancer cells develop resistance to targeted therapies. We believe that inhibition of SHP2 could block a common path that cancer cells exploit to avoid killing by other antitumor agents, thus overcoming or delaying the onset of resistance to those therapies. We are currently evaluating the safety and tolerability of RLY-1971 in a Phase 1 dose escalation study in patients with advanced or metastatic solid tumors. In December 2020, we entered into the Genentech Agreement, a global collaboration and license agreement with Genentech for the clinical development and commercialization of RLY-1971. Given the range of cancers that are related to SHP2 dependence, we believe RLY-1971 has the potential to serve as a combination backbone therapy.

We estimate there are approximately 55,000 late-line patients annually in the United States with advanced lung cancer who might benefit from a combination of RLY-1971 with another targeted inhibitor. In the future, if RLY-1971 advances to earlier lines of combination treatment for lung cancer, we believe it could be applied in the treatment of approximately 90,000 patients annually in the United States. The subset of patients with KRAS G12C mutations in lung cancer that could potentially benefit from the combination of RLY-1971 with GDC-6036 is approximately 15,000-25,000 annually in the United States.

SHP2: a central regulator of cell signaling

SHP2 is a protein tyrosine phosphatase that plays a critical role in the transduction of intracellular signals downstream from RTKs, promoting cell survival and growth through the RAS pathway. SHP2 was the first phosphatase identified as a recurrently mutated oncogene, providing genetic support for the importance of SHP2 activation in promoting cancer. In addition to the central role of SHP2 in RTK signaling, some alterations in the RAS signaling pathway amplify signals transmitted by SHP2 and can therefore be suppressed by SHP2 inhibition. These include specific mutant forms of RAS (KRAS G12C and KRAS G12A), genomic amplification of wild-type KRAS, loss-of-function mutations in NF1, and class 3 mutations in BRAF. Consequently, there are multiple cancer genetic contexts where SHP2 inhibition could be beneficial as a monotherapy.

A key feature of SHP2 as an oncology target is its ability to regulate cell signaling that arises from multiple RTKs (Figure 7). Therapies targeted to these RTKs, and therapies targeting downstream nodes such as PI3K, KRAS and MEK, are often unable to durably inhibit tumor growth because these tumors are able to bypass the targeted RTK and shift growth factor signaling to an alternate RTK, rendering them less sensitive to the targeted therapy. This is generally referred to as bypass resistance. Because SHP2 regulates the activity of multiple RTKs, inhibition of SHP2 is an effective way to overcome bypass resistance as confirmed by cellular and animal model experiments. Indeed, added benefit of SHP2 inhibition has been demonstrated pre-clinically in combination with multiple agents, such as those targeting MEK, KRASG12C, EGFR, and ALK. We believe our SHP2 inhibitor has the potential to become a commonly used combination partner with multiple targeted therapies including those in our own pipeline.

Figure 7: SHP2 regulates the activity of multiple receptor tyrosine kinases (RTKs).

Our solution, RLY-1971

RLY-1971 is a small molecule inhibitor of SHP2 that binds and stabilizes SHP2 in its inactive conformation.

We utilized a combination of experimental and computational techniques to identify unique inhibitors. For example, using long timescale MD simulations we were able to understand changes in the dynamics of the binding pocket over time that would not have been appreciated with shorter timescale simulations (Figure 8). Informed by high-resolution room-temperature X-ray crystallographic data, we created a virtual representation of our lead molecule bound to the SHP2 protein. We then simulated this system over long timescales. As shown in Figure 8, we observed that a loop (green) to the left of the small molecule (orange) moves down towards the molecule over the course of the simulation. Our medicinal chemists were then able to leverage this understanding in their designs to create an inhibitor of SHP2. Importantly, this loop cannot be resolved using conventional X-ray crystallography. Therefore, relying on standard techniques could deprive medicinal chemists of a critical insight as they attempt to design improved compounds.

Figure 8: We depict a small molecule docked in a representation of the SHP2 protein where there is a green loop visible to the left of the small molecule (orange). A 500 ns MD simulation (0.5 μs) shows that the green loop is far away from the small molecule (left). A longer simulation (10.0 μs), reveals that the loop flips downwards, close to where the small molecule binds (right).

We then prioritized compounds with the best predicted binding to SHP2 over a 10 μs molecular dynamics simulation and tested the most stable compounds in our biochemical assay. This enabled filtering and prioritization of candidate molecules, resulting in the identification of RLY-1971, our clinical-stage compound. RLY-1971 inhibits SHP2 phosphatase activity (750 pM IC50) in a biochemical assay designed to monitor dephosphorylation of a probe substrate. RLY-1971 also inhibits SHP2 in cellular assays, as measured by inhibition of ERK1/2 phosphorylation at Thr202/Tyr204 (1.3 nM IC50 in KYSE-520, an EGFR amplified gastric cancer cell line), and by inhibition of cancer cell proliferation (70 nM IC50 in KYSE-520 and 11 nM IC50 in NCI-H358, a KRASG12C mutant NSCLC cell line) (Figure 9).

Figure 9: RLY-1971 potently inhibits SHP2 in biochemical and cellular assays.

RLY-1971 shows minimal inhibition of targets other than SHP2. RLY-1971 has bioavailability suitable for oral dosing, is metabolically stable, and demonstrates favorable pharmacokinetic properties in preclinical in vivo models. We do not predict that RLY-1971 will have significant drug-drug interactions based on weak inhibition of drug metabolizing enzymes. It is readily synthesized in bulk, can be formulated for oral delivery, and was well-tolerated in animal models.

We believe the key differentiating features of RLY-1971 from other SHP2 inhibitors in clinical development are:

RLY-1971 monotherapy pre-clinical experience

SHP2 inhibition has been shown in third party studies to result in tumor stasis or regression in preclinical xenograft models of tumors harboring KRAS genomic amplification, KRASG12C mutations, NF1LOF mutations, or BRAFClass3 mutations. Consistent with these findings, in our internal pre-clinical studies, RLY-1971 inhibited the proliferation of a panel of cancer cell lines driven by KRAS mutations that require signals transmitted by SHP2 (KRASG12C and KRASG12A) but was inactive in cancer cell lines driven by other KRAS mutations that do not require SHP2 signals (KRASG12D) (Figure 10).

Figure 10: Inhibition of proliferation by RLY-1971 in cancer cell lines driven by KRAS mutations. Mutations that require SHP2 signals are potently inhibited by RLY-1971, whereas mutations that do not require SHP2 signals are insensitive to RLY-1971.

A panel of KRASG12 mutant cancer cell lines were grown in 3D spheroids and treated with RLY-1971 in a proliferation assay. KRASG12C and KRASG12A mutations retain intrinsic GTPase activity and therefore require SHP2 signaling, whereas the KRASG12D mutation does not. For cell lines indicated, cells were plated at a density of 2000 cells/well in round bottom ultra-low attachment 384-well plates (Corning) in growth media and cells are allowed to form three-dimensional structures at 37°C, 5% CO2 incubator for 48 hours. After a 48 hour incubation period, cells were then treated in triplicate with serial 3-fold dilutions of inhibitor in complete growth media and cells were returned to incubator for an additional 120 hours. CellTiter-Glo 3D reagent (Promega) was then added into each well and incubated are room temperature for 30 minutes followed by reading on an EnVision Reader (Perkin Elmer) using standard conditions. Assay data was normalized to DMSO control wells. Dose response curve fitting and IC50 values were determined using Genedata analyzer.

To demonstrate activity of RLY-1971 as a single agent in vivo, we tested it in multiple cancer xenograft mouse models. Consistent with our in vitro data and the role of SHP2 as a critical mediator of RTK signaling, we observed that RLY-1971 induced regression in cancer xenograft models harboring a KRASG12C mutation or genomic amplification of EGFR when administered on a continuous dosing schedule (Figure 11).

Figure 11: RLY-1971 induces regression in KRAS G12C mutant and EGFR amplified cancer xenograft models.

Anti-tumor activity of the SHP2 inhibitor RLY-1971 dosed twice daily by oral administration (PO BID) in (a)the KRAS G12C mutant NSCLC xenograft model NCIH358 after 28 days on treatment and (b)the EGFR amplified gastric cancer xenograft model KYSE-520 after 21 days on treatment. Treatment with RLY-1971 resulted in dose-dependent anti-tumor activity and regression in both models. Data represent waterfall plots of individual end of study tumors, with tumor volume expressed as percentage change relative to initial tumor volume. Each animal is represented as a separate bar (number of mice per group = 8). For each of the studies the statistical difference between the vehicle treated and RLY-1971 treated groups was assessed by one-way analysis of variance (ANOVA). In both studies the RLY-1971 treated groups were determined to be significantly different than the vehicle treated group with a P value < 0.001.

In addition, RLY-1971 inhibited the proliferation of cancer cell lines engineered to express known cancer mutations in SHP2 (Figure 12). These mutations bias SHP2 towards an open, active conformation in direct opposition to the allosteric inhibition effected by RLY-1971. RLY-1971 retains nanomolar potency against activating mutations of SHP2. We hypothesize that the activity of RLY-1971 against activating mutations of SHP2 could result in more durable benefit by suppressing the emergence of resistant cell populations with SHP2 resistance mutations.

Figure 12: RLY-1971 inhibits the proliferation of cells expressing known SHP2 activating mutations.

Inhibition of proliferation by RLY-1971 in TF1 cancer cells expressing known cancer mutations in SHP2. TF1 cells were engineered to express the SHP2 mutations (a)E69K (IC50 = 18.4 nM) or (b)A461T (IC50 = 7 nM) and

treated with RLY-1971 in a proliferation assay. 500 cells/well were seeded in round bottom ultra-low attachment 384-well plates (Corning) in growth media and incubated for 48 hours at 37°C in 5% CO2. Cells were then treated in triplicate with serial 3-fold dilutions of inhibitor in growth media. Following incubation in the presence of compound for an additional 120 hours, cell viability was determined using the CellTiter-Glo 3D assay kit (Promega) following the manufacturer’s instructions. Luminescence was read in an EnVision Multimode Plate Reader (Perkin Elmer). Assay data was normalized to DMSO values, and dose response curve fitting was performed using Genedata analyzer.

RLY-1971 as a combination therapy

Given the role of SHP2 in mediating bypass resistance, we believe that SHP2 inhibitors have significant therapeutic potential when given in combination with other targeted therapies. Due to the increased potency and broader mutational coverage of next-generation targeted therapies, lower rates of on-target resistance have been observed in the clinic, with a greater number of patients progressing due to bypass resistance. An example of this is seen with EGFR inhibitors, where first-generation inhibitors (erlotinib and gefinitib) have greater on-target resistance compared to a third-generation inhibitor (osimertinib). As SHP2 is involved in signaling for numerous oncogenes, including EGFR, KRASG12C, ALK and MET, combination therapy with RLY-1971 represents a potential significant therapeutic opportunity.

Consistent with the role of SHP2 in RTK signaling in NSCLC, in our pre-clinical experiments, RLY-1971 demonstrated combination benefit in cell culture experiments when co-administered with inhibitors of MEK, ALK, or EGFR.

To demonstrate combination benefit with our SHP2 inhibitor in vivo, we combined RLY-1971 with the ALK inhibitor alectinib in an ALK-translocated NSCLC xenograft mouse model (NCIH3122) that was derived in vitro to have reduced sensitivity to ALK inhibition (Figure 13). DNA sequencing did not reveal new ALK mutations in the cell line. Therefore, these cells likely have reduced sensitivity due to a bypass mechanism. The combination of RLY-1971 with alectinib resulted in tumor regressions in all treated animals.

Figure 13: Anti-tumor activity of RLY-1971 and the ALK inhibitor alectinib as single agents or in combination in an ALK translocated NSCLC xenograft model (NCI-H3122) derived in vitro to have reduced sensitivity to ALK inhibition.

Daily oral administration of RLY-1971 at 30mpk BID in combination with alectinib at 10mpk QD (green) resulted in increased efficacy compared to alectinib at 10mpk QD (red) or RLY-1971 at 30mpk BID (orange) alone in an ALK translocated NSCLC xenograft model (NCI-H3122) derived in vitro to be less sensitive to ALK inhibition. Data represent waterfall plots of individual tumors after 27 days of treatment with compounds, with tumor volume expressed as percentage change relative to initial tumor volume. Each animal is represented as a separate bar

(number of mice per group = 9). The statistical difference between the combination treated group and the RLY-1971 or alectinib single agent groups was assessed using an unpaired t-test. The combination treated group with significant with P-value <0.001 compared to either single agent group.

In addition to RTK inhibitors, combination benefit for SHP2 inhibition has been demonstrated with other targeted agents including MEK inhibitors and KRASG12C inhibitors in cancer xenograft models harboring KRASG12C mutations or KRAS amplifications. The efficacy of direct KRASG12C inhibition may be limited by adaptive feedback reactivation of the RAS-MAPK pathway through upregulation of multiple RTKs. Activation of these RTKs leads to compensatory activation of wild-type RAS isoforms, which cannot be inhibited by KRASG12C-specific inhibitors, thus leading to resistance. SHP2 is unique in that it transmits signals from multiple RTKs and is therefore critical in mediating feedback reactivation of the RAS pathway during KRASG12C inhibition.

Consistent with these observations, RLY-1971 demonstrated in vivo combination benefit with the KRASG12C specific inhibitor AMG-510 in a KRASG12C lung cancer xenograft model (Figure 14). Specifically, the combination resulted in regression in all animals, whereas each single agent resulted in more modest activity at the doses that were tested. These results suggest that SHP2 inhibition abrogates compensatory RAS-MAPK pathway activation during KRASG12C inhibition. Molecular characterization of phosphorylated-ERK, or Perk, a downstream marker of RAS-MAPK pathway activity, supports this conclusion. In vitro, the combination of RLY-1971 and the KRASG12C-specific inhibitor ARS-1620 was able to fully suppress pERK in this model, while each inhibitor individually only partially suppressed pERK. Based on these data, we believe that the combination of RLY-1971 with KRASG12C-specific inhibitors warrants clinical studies in patients with tumors harboring KRASG12C mutations.

Figure 14: RLY-1971 and the KRASG12C-specific inhibitor AMG-510 demonstrate synergy when used in combination in the KRASG12C NCI-H358 lung cancer cell line.

In vivo combination benefit of RLY-1971 and the KRASG12C-specific inhibitor AMG-510. Anti-tumor activity of the SHP2 inhibitor RLY-1971 dosed twice daily (BID) by oral administration at 30 mpk and the KRAS G12C inhibitor AMG-510 dosed once daily (QD) by oral administration at 10 mpk in the KRAS G12C mutant NSCLC xenograft model NCIH358. Treatment with the combination resulted in regression in all animals. Data represent waterfall plots of individual end of study tumors, with tumor volume expressed as percentage change relative to initial tumor volume. Each animal is represented as a separate bar (number of mice per group = 8)

In addition to the therapeutic opportunity associated with combining with other targeted therapies, we believe RLY-1971 has the potential to be a combination partner with the product candidates in our own precision oncology portfolio, RLY-4008 and RLY-PI3K1047.

Our clinical development plan

In the first quarter of 2020, we began evaluating the safety and tolerability of RLY-1971 in a first-in-human dose escalation study in patients with advanced or metastatic solid tumors. The primary objectives are to determine the maximum tolerated dose (MTD)/recommended phase 2 dose (RP2D), and to define the overall safety profile of RLY-1971. Secondary objectives are to assess the pharmacokinetics, pharmacodynamics, and to explore preliminary anti-tumor activity of RLY-1971. Patients will receive RLY-1971 administered orally, once daily. Once daily oral dosing was selected based on projected human pharmacokinetics and exposures calculated from multi-species pharmacokinetics and allometric scaling.

The first-in-human monotherapy data will facilitate subsequent clinical evaluation and development of RLY-1971 in combination with other targeted therapies in indications where SHP2 inhibition may exert synergistic antitumor effects. Future development for RLY-1971 will be governed by a joint development team between us and Genentech. We expect a combination trial of RLY-1971 and Genentech’s KRAC G12C inhibitor, GDC-6036, to be initiated in 2021.

Our Discovery Programs

We are deploying our Dynamo platform to advance an additional three discovery-stage precision oncology programs. As with our lead programs, these programs leverage insights into protein conformational dynamics to address high-value, genetically validated oncogenes that previously have been intractable to conventional drug-discovery approaches. The capabilities for our Dynamo platform in protein visualization can be applied to multiple therapeutic areas beyond precision oncology. We are continuing to leverage the power of our Dynamo platform to further diversify our pipeline by extending our approach to address genetically validated targets in monogenic diseases with two discovery-stage programs, where genetic alterations lead to disease-causing defects in protein conformational dynamics.

Competition

The biotechnology and pharmaceutical industries are characterized by rapid innovation of new technologies, fierce competition and strong defense of intellectual property. While we believe that our platform and our knowledge, experience and scientific resources provide us with competitive advantages, we face competition from major pharmaceutical and biotechnology companies, academic institutions, governmental agencies and public and private research institutions, among others.

We compete in the segments of the pharmaceutical, biotechnology, and other related markets that address experimentally and computationally driven structure-based drug design in cancer and genetic diseases. There are other companies focusing on structure-based drug design to develop therapies in the fields of cancer and other diseases. These companies include divisions of large pharmaceutical companies and biotechnology companies of various sizes. Any product candidates that we successfully develop and commercialize will compete with currently approved therapies and new therapies that may become available in the future from segments of the pharmaceutical, biotechnology and other related markets that pursue precision medicines. Key product features that would affect our ability to effectively compete with other therapeutics include the efficacy, safety and convenience of our products.

We believe principal competitive factors to our business include, among other things, the rich protein structural data sets we are able to generate, the power and accuracy of our computations and predictions, ability to integrate experimental and computational capabilities, ability to successfully transition research programs into clinical development, ability to raise capital, and the scalability of the platform, pipeline, and business.

While there are many pharmaceutical and biotechnology companies that use some of the same tools that we use in our platform, we believe we compete favorably on the basis of these factors. The effort and investment required to develop a highly integrated experimental and computational platform similar to ours will hinder new entrants that are unable to invest the necessary capital and time and lack the breadth and depth of technical expertise required to develop competing capabilities. Our ability to remain competitive will largely depend on our ability to continue to augment our integrated experimental and computational platform and demonstrate success in our drug discovery efforts.

Our competitors may obtain regulatory approval of their products more rapidly than we may or may obtain patent protection or other intellectual property rights that limit our ability to develop or commercialize our product candidates. Our competitors may also develop drugs that are more effective, more convenient, more widely used and less costly or have a better safety profile than our products and these competitors may also be more successful than us in manufacturing and marketing their products.

In addition, we will need to develop our product candidates in collaboration with diagnostic companies, and we will face competition from other companies in establishing these collaborations. Our competitors will also compete with us in recruiting and retaining qualified scientific, management and commercial personnel, establishing clinical trial sites and patient registration for clinical trials, as well as in acquiring technologies complementary to, or necessary for, our programs.

Furthermore, we also face competition more broadly across the market for cost-effective and reimbursable cancer treatments. The most common methods of treating patients with cancer are surgery, radiation and drug therapy, including chemotherapy, hormone therapy and targeted drug therapy or a combination of such methods. There are a variety of available drug therapies marketed for cancer. In many cases, these drugs are administered in combination to enhance efficacy. While our product candidates, if any are approved, may compete with these existing drug and other therapies, to the extent they are ultimately used in combination with or as an adjunct to these therapies, our product candidates may not be competitive with them. Some of these drugs are branded and subject to patent protection, and others are available on a generic basis. Insurers and other third-party payors may also encourage the use of generic products or specific branded products. We expect that if our product candidates are approved, they will be priced at a significant premium over competitive generic, including branded generic, products. As a result, obtaining market acceptance of, and gaining significant share of the market for, any of our product candidates that we successfully introduce to the market will pose challenges. In addition, many companies are developing new therapeutics, and we cannot predict what the standard of care will be as our product candidates progress through clinical development.

RLY-4008

While there are currently no approved products that selectively target FGFR2, we are aware of other companies developing therapeutics that selectively target FGFR2, including, but not limited to, Five Prime Therapeutics and Russian Pharmaceutical Technologies. Specifically, we expect RLY-4008 to compete with approved development stage non-selective inhibitors of the FGFR receptor family that are being tested in patients with FGFR2 alterations, including but not limited to, Incyte Corporation (pemigatinib), QED Therapeutics (infigratinib), Basilea Pharmaceutica AG (derazantinib), Janssen Pharmaceuticals, Inc. (erdafitinib), Otsuka Holdings Co., Ltd. through its subsidiary Taiho Pharmaceutical Co., Ltd. (TAS-120), Debiopharm Group (Debio1347), Eisai Co., Ltd. (E-7090), and InnoCare Pharma Limited (ICP-192).

The development of RLY-4008 will focus on solid tumor patients with FGFR2 alterations, including intrahepatic cholangiocarcinoma (ICC) patients harboring FGFR2 gene fusions. While there are no approved systemic therapies for ICC, the current standard of care for unresectable or metastatic patients is first-line gemcitabine/cisplatin chemotherapy. In addition, there are other companies developing potentially competitive drug candidates in ICC including, but not limited to, Merck & Co, Astrazeneca plc, Merck KGaA, and NuCana plc.

Mutant-PI3Kα Inhibitor Program

We expect that our mutant-selective PI3Kα inhibitor program will compete against an approved drug, Piqray (alpelisib), a non-selective PI3Kα inhibitor marketed by Novartis for the treatment of PI3Kα mutated breast cancer.

We are aware of other companies developing therapeutics that target both wild-type and mutant PI3Kα, including, but not limited to, Roche Holding AG through its subsidiary Genentech, Petra Pharma Corporation, Menarini Group, Luoxin Pharma and Shanghai HaiHe Pharma Co. Petra Pharma Corporation also has a preclinical development program for a mutant-selective PI3Kα inhibitor.

RLY-1971

While there are currently no approved products targeting SHP2, we are aware of other companies in clinical trials developing therapeutics that target SHP2, including, but not limited to, Revolution Medicines in partnership with Sanofi S.A., Novartis International AG, Navire Pharma, Inc., Erasca, Inc. and Jacobio Pharmaceuticals, Inc. in partnership with AbbVie Inc.

Our Collaborations

Key Scientific Collaborations

While we have invested extensively in our in-house capabilities and know-how, we selectively work with key collaborators and field experts on certain emerging technologies. Most of our experimental collaborations are focused on the technologies we use to visualize protein structure at the atomic level. For example, we work with Professor James Fraser from UCSF on performing and analyzing room temperature X-ray crystallography experiments and Professor Adam Frost from UCSF on Cryo-EM image analysis. Both are world leading experts on these technologies, and they provide important know-how and insights in collaboration with our scientists.

Our key computational collaboration is with D. E. Shaw Research, LLC, or D. E. Shaw Research, a computational biochemistry research firm operating under the scientific leadership of Dr. David E. Shaw, which has developed proprietary software and hardware to perform long timescale molecular dynamics simulations. Through an affiliate, D. E. Shaw Research is also one of our investors. We collaborate with D. E. Shaw Research scientists to research certain protein targets on an exclusive basis, with a focus on the dynamic behavior of proteins, through the use of D. E. Shaw Research’s computational modeling capabilities, such as the Anton 2 supercomputer and proprietary algorithms and software developed specifically by D. E. Shaw Research for processing long timescale molecular dynamics simulations. Our scientists work closely with D. E. Shaw Research scientists on each of our programs, especially in the discovery stage as we develop motion-based hypotheses and identify lead compounds. See “—License Agreements and Strategic Collaborations —Collaboration and License Agreement with D. E. Shaw Research, LLC” for more detail on the terms of the DESRES Agreement.

We also have other collaborations mostly focused on developing machine learning models. Specifically, we collaborate with Google on machine learning models to generate novel molecules with specific activity, and with Professor Tim Cernak from the University of Michigan on machine learning models focused on chemical synthesis and high throughput experimentation.

License Agreements and Strategic Collaborations

Collaboration and License Agreement with D. E. Shaw Research, LLC

On June 15, 2020, we entered into an Amended and Restated Collaboration and License Agreement with D. E. Shaw Research, extending the term and otherwise modifying the terms of a Collaboration and License Agreement originally entered into on August 17, 2016, as amended. We refer to this amended and restated agreement as the DESRES Agreement. Under the DESRES Agreement, we agreed to collaborate with D. E. Shaw Research to research certain biological targets through the use of D. E. Shaw Research computational modeling capabilities focused on analysis of protein motion, with an aim to develop and commercialize compounds and products directed to such targets. After completing the computational modeling with D. E. Shaw Research and naming a compound development candidate, we develop and commercialize such compounds and products. D. E. Shaw Research has no

involvement with the clinical development or potential commercialization of these compounds and products, regardless of any co-ownership rights pursuant to the terms of the DESRES Agreement, and instead receives solely milestone and royalty payments as described below.

Under the DESRES Agreement, there are three categories of targets: Category 1 Targets, Category 2 Targets and Category 3 Targets. We and D. E. Shaw Research agreed on a list of Category 1 Targets and Category 2 Targets as part of the DESRES Agreement. Category 1 Targets are targets that, among other things, we collaborate on with D. E. Shaw Research, D. E. Shaw Research has exclusivity obligations with respect to, and we may owe royalties on; Category 2 Targets are targets in connection with the potential re-categorization of which into a Category 1 Target, we may, among other things, perform in vitro non-clinical research and development (but not in vivo non-clinical development, clinical development or commercialization), and Category 3 Targets are all targets other than Category 1 Targets and Category 2 Targets. There are mechanisms for re-categorizing targets, and we and D. E. Shaw Research have re-categorized a number of targets during the first four years of our collaboration. Our rights and obligations, and D. E. Shaw Research’s rights and obligations, with respect to targets vary by the category of each target. However, the parties only conduct collaborative activities together for Category 1 Targets, and we are limited to a maximum of eleven Category 1 Targets in the current collaboration year (with such number potentially changing from year to year, with any increase in such number of targets subject to the collaboration in each collaboration year capped at four more than the highest number of such targets in the previous year). The sum of the number of Category 1 Targets and the number of Category 2 Targets is capped at twenty, in any event.

Work product that is jointly developed with D. E. Shaw Research is initially co-owned with them. Specifically, intellectual property rights covering the composition of matter for RLY-1971 are currently co-owned by D. E. Shaw Research and us under this arrangement. We have the right to have patents claiming certain product candidates (including one claiming RLY-1971) assigned to us upon issuance of those patents. Although other compounds in our FGFR2 and PI3Kα programs were jointly conceived with D. E. Shaw Research, RLY-4008 and RLY-PI3K1047 were conceived solely by Relay Therapeutics inventors. For each Category 1 Target there is a limit of up to 10 core compounds and a total of 500 compounds including derivatives of those core compounds that can be designated as solely owned by us, provided that if D. E. Shaw Research provides us with notice that certain compounds cannot be designated as solely owned by us due to concerns in respect of a Category 3 Target, then the limit on Category 1 Target core compounds will increase by one and the limit on total compounds will increase by fifty, but subject to a maximum of 15 and 750, respectively, for each Category 1 Target. Each of we and D. E. Shaw Research grants to the other a perpetual, irrevocable, non-exclusive license for jointly held intellectual property, subject to certain exclusions.

During the initial research term, which is expected to last until August 2025, unless extended by mutual agreement, D. E. Shaw Research will not, and will cause its subsidiaries not to, research any Category 1 Target (or grant certain rights with respect to such target) with the aim of pursuing any compound designed to interact with or bind to such Category 1 Target, subject to some exceptions. Following the end of the initial research term, D. E. Shaw Research will not, and will cause its subsidiaries not to, research a Category 1 Target (or grant certain rights with respect to such target) with the aim of pursuing any compound designed to interact with or bind to any target that was a Category 1 Target at the end of the initial research term, subject to some exceptions. D. E. Shaw Research will not be bound by such exclusivity provisions with respect to a particular Category 1 Target if we, and parties acting on our behalf, stop using commercially reasonable efforts to research, develop or commercialize any products against such Category 1 Target. Further, D. E. Shaw Research will be released from such exclusivity obligations with respect to a particular Category 1 Target if, at least 24 months after the end of the initial research term, D. E. Shaw Research informs us that D. E. Shaw Research will forgo all future payments with respect to such Category 1 Target.

During the initial research term, neither D. E. Shaw Research nor we will, and we will each cause our subsidiaries not to, research a Category 2 Target (or grant certain rights with respect to such target) with the aim of pursuing any compound designed to interact with or bind to such Category 2 Target, subject to some exceptions. These exclusivity restrictions do not extend past the initial research term.

There is no exclusivity with respect to Category 3 Targets.

Through December 31, 2020, we have made cash payments to D. E. Shaw Research totaling $8.4 million in the aggregate. On a product-by-product basis, we have also agreed to pay D. E. Shaw Research milestone payments upon the achievement of certain development and regulatory milestone events for products we develop under the DESRES Agreement that are directed to a Category 1 Target or any target that was a Category 1 Target. Our SHP2, FGFR2 and PI3K programs are each directed to Category 1 Targets. Such payments for achievement of development and regulatory milestones total up to $7.25 million in the aggregate for each of the first three products we develop, and up to $6.25 million in the aggregate for each product we develop after the first three.

Additionally, we have agreed to pay D. E. Shaw Research, on a product-by-product basis, with respect to products directed to Category 1 Targets or any target that was a Category 1 Target, royalties in the low single digits on worldwide net sales of products that we commercialize directed to the targets selected for development under the DESRES Agreement, subject to certain reductions. Royalties are payable on a product-by-product and country-by-country basis until the later of twelve years after first commercial sale in such country or the expiration of all applicable regulatory exclusivities in such country. On a product-by-product basis, we also agreed to pay D. E. Shaw Research sales milestone payments up to $36.0 million in the aggregate based on sales of each product directed to a Category 1 Target or any target that was a Category 1 Target. Further, if we enter into transactions granting third parties rights to a Category 1 Target or a compound or product directed to a Category 1 Target or any target that was a Category 1 Target.such as our collaboration with Genentech for RLY-1971 discussed below, but subject to certain exclusions, we will share with D. E. Shaw Research a percentage of the proceeds of such transactions ranging from the low- to high-single digits, depending on the stage of development of compounds or products directed to such target at the time we enter into such transaction. We have also agreed to pay D. E. Shaw Research an annual collaboration fee in August of each year during the initial research term, such fee to be $7,900,000 for each year between 2020 and 2025.

Unless earlier terminated, the DESRES Agreement will continue at least until the end of the initial research term and thereafter on a target-by-target basis until all payment obligations have expired. D. E. Shaw Research has the right to terminate the DESRES Agreement due to non-payment. We and D. E. Shaw Research each have the right to terminate the DESRES Agreement due to an uncured material breach by the other party, or in the event the other party becomes insolvent or enters into bankruptcy or dissolution proceedings. Our payment obligations to D. E. Shaw Research survive termination of the DESRES Agreement. If D. E. Shaw Research terminates the DESRES Agreement, the exclusivity obligations will terminate. If we terminate the DESRES Agreement, D. E. Shaw Research remains bound by its exclusivity obligations with respect to certain targets until, on a target-by-target basis, there are no further payment obligations due to D. E. Shaw Research in respect of such targets.

Collaboration and License Agreement with Genentech

On December 11, 2020, we entered into a Collaboration and License Agreement with Genentech, Inc. and F. Hoffmann-La Roche Ltd, collectively referred to as Genentech. We refer to this agreement as the Genentech Agreement. Pursuant to the Genentech Agreement, we and Genentech will collaborate on the development and commercialization of RLY-1971.

Unless Genentech elects to exercise its option to conduct the remainder of the ongoing Phase 1a clinical trial for RLY-1971, we will complete this trial. Genentech will be responsible for conducting all subsequent clinical development of RLY-1971, including in any combination trials with Genentech’s compound, GDC-6036, that directly binds to and inhibits KRAS G12C, or other compounds.

We retain the right to develop RLY-1971 or certain other small molecule inhibitors of SHP2 developed under the Genentech Agreement, or a Licensed Candidate, or pharmaceutical product containing a Licensed Candidate, or a Licensed Product, in combination with any of our compounds targeting FGFR2, including RLY-4008, or PI3Kα, including candidates in our RLY-PI3K1047 program, which we refer to as a Relay Combination Product. If we opt into the Profit/Cost Share described below, Genentech may share the development costs of any clinical trial for a Relay Combination Product.

Genentech has the sole right and responsibility to commercialize Licensed Products, in any and all combinations, except that we have the right to co-promote a Licensed Product solely as part of our commercialization of Relay Combination Products. Genentech will be solely responsible for all regulatory matters for all Licensed Candidates

and Licensed Products after the assignment by us to Genentech of all related regulatory materials, including the IND application for the Phase 1a Trial, other than with respect to Relay Combination Products.

Under the terms of the Genentech Agreement, we have received $75 million in an upfront payment and are eligible to receive $25 million in additional near-term payments.

We have the option, exercisable one time in our sole discretion, to fund half of the development costs of RLY-1971 in the United States and share half of the net profits or net loss of commercializing RLY-1971 in the United States, which we refer to as the Profit/Cost Share. If we opt into the Profit/Cost Share, we will also be eligible to receive up to an aggregate of an additional $410 million upon the achievement of specified commercialization and sales-based milestones for RLY-1971 outside of the United States and tiered royalties ranging from low-to-mid teens on annual net sales of RLY-1971 outside of the United States, on a country-by-country basis, subject to reduction in certain circumstances. At any time prior to the third anniversary of the first commercial sale of RLY-1971 in the United States, we may elect to opt-out of further participation in the Profit/Cost Share. If we elect to opt-out, then Genentech’s milestone and royalty payment obligations will revert to the financial terms that would be applicable if we had not opted into the Profit/Cost Share as described below as of the effective opt-out date, with certain adjustments.

If we do not opt into the Profit/Cost Share, Genentech will be responsible for all development costs of RLY-1971 other than the costs incurred by us for the ongoing Phase 1a trial of RLY-1971, and we will be eligible to receive up to an aggregate of an additional $695 million upon the achievement of specified development, commercialization and sales-based milestones for RLY-1971 worldwide. We will also be eligible to receive tiered royalties ranging from low-to-mid teens on annual worldwide net sales of RLY-1971, on a country-by-country basis, subject to reduction in certain circumstances.

In the event of regulatory approval of both RLY-1971 and GDC-6036 in combination, we are eligible to receive additional royalties.

Under the Genentech Agreement, we granted an exclusive, worldwide, royalty-bearing license to Genentech, with the right to sublicense, to develop and commercialize RLY-1971. Between the parties, Genentech has the first right, but not the obligation, to file, prosecute and maintain any patents licensed to it pursuant to the Genentech Agreement, as well as to enforce infringement of or defend claims against such patents that relate to Licensed Candidates and Licensed Products. The parties will share any liabilities or damages arising from the enforcement of such patents or any third-party patent claims.

Other than with respect to Relay Combination Products and other activities in accordance with the Genentech Agreement, we may not, directly or indirectly, conduct any activities related to the research, development, manufacture or commercialization of any SHP2 inhibitor. During the first three years of the term of the Genentech Agreement, Genentech will cause its research and early development organization not to sponsor or conduct a registrational trial for a SHP2 inhibitor other than a Licensed Product.

Unless earlier terminated, the Genentech Agreement will remain in effect until the later of the date on which Genentech is no longer developing or commercializing RLY-1971 in the United States if we have opted into the Profit/Cost Share and have not subsequently opted-out, or the expiration of all Genentech’s royalty payment obligations to us. The parties may terminate the Genentech Agreement for the other party’s material breach or insolvency or, on a country-by-country basis, the failure to obtain merger control under applicable antitrust laws. Additionally, Genentech may terminate the Genentech Agreement for convenience, and we may terminate the Genentech Agreement for certain patent challenges by Genentech or if Genentech has not conducted any research, development, manufacturing or commercialization activities with respect to any Licensed Candidate or Licensed Product for a specified period.

Intellectual Property

We seek to protect the intellectual property and proprietary technology that we consider important to our business, including by pursuing patent applications that cover our product candidates and methods of using the same, as well as any other relevant inventions and improvements that we believe to be commercially important to the development of our business. We also rely on trade secrets, know-how and continuing technological innovation to develop and maintain our proprietary and intellectual property position. Our commercial success depends, in part, on our ability

to obtain, maintain, enforce and protect our intellectual property and other proprietary rights for the technology, inventions and improvements we consider important to our business, and to defend any patents we may own or in-license in the future, prevent others from infringing any patents we may own or in-license in the future, preserve the confidentiality of our trade secrets, and operate without infringing, misappropriating or otherwise violating the valid and enforceable patents and proprietary rights of third parties.

As with other biotechnology and pharmaceutical companies, our ability to maintain and solidify our proprietary and intellectual property position for our product candidates and technologies will depend on our success in obtaining effective patent claims and enforcing those claims if granted. However, our pending provisional and PCT patent applications, and any patent applications that we may in the future file or license from third parties, may not result in the issuance of patents and any issued patents we may obtain do not guarantee us the right to practice our technology or commercialize our product candidates. We also cannot predict the breadth of claims that may be allowed or enforced in any patents we may own or in-license in the future. Any issued patents that we currently own or may own or in-license in the future may be challenged, invalidated, circumvented or have the scope of their claims narrowed. In addition, because of the extensive time required for clinical development and regulatory review of a product candidate we may develop, it is possible that, before any of our product candidates can be commercialized, any related patent may expire or remain in force for only a short period following commercialization, thereby limiting the protection such patent would afford the respective product and any competitive advantage such patent may provide.

The term of individual patents depends upon the date of filing of the patent application, the date of patent issuance and the legal term of patents in the countries in which they are obtained. In most countries, including the United States, the patent term is 20 years from the earliest filing date of a non-provisional patent application. In the United States, a patent’s term may be lengthened by patent term adjustment, which compensates a patentee for administrative delays by the USPTO in examining and granting a patent, or may be shortened if a patent is terminally disclaimed over an earlier expiring patent. The term of a patent claiming a new drug product may also be eligible for a limited patent term extension when FDA approval is granted, provided statutory and regulatory requirements are met. The restoration period granted on a patent covering a product is typically one-half the time between the effective date of a clinical investigation involving human beings is begun and the submission date of an application, plus the time between the submission date of an application and the ultimate approval date. The restoration period cannot be longer than five years, and the restoration period may not extend the patent term beyond 14 years from the date of FDA approval. Only one patent applicable to an approved product is eligible for the extension, and only those claims covering the approved product, a method for using it, or a method for manufacturing it may be extended. Additionally, the application for the extension must be submitted prior to the expiration of the patent in question. A patent that covers multiple products for which approval is sought can only be extended in connection with one of the approvals. The United States Patent and Trademark Office reviews and approves the application for any patent term extension or restoration in consultation with the FDA. In the future, if our product candidates receive approval by the FDA, we expect to apply for patent term extensions on one issued patent covering each of those products, depending upon the length of the clinical studies for each product and other factors. There can be no assurance that patents will issue from our current or future pending patent applications, or that we will benefit from any patent term extension or favorable adjustments to the terms of any patents we may own or in-license in the future. In addition, the actual protection afforded by a patent varies on a product-by-product basis, from country-to-country, and depends upon many factors, including the type of patent, the scope of its coverage, the availability of regulatory-related extensions, the availability of legal remedies in a particular country and the validity and enforceability of the patent. The patent term may be inadequate to protect our competitive position on our products for an adequate amount of time.

RLY-4008

As of February 28, 2021, we co-owned with D. E. Shaw Research pending PCT, Argentine, and Taiwanese applications which relate to our FGFR2 inhibitors. Any U.S. or foreign patent that may issue from these patent applications would be scheduled to expire in 2040, excluding any additional term for patent term adjustment or patent term extension, if applicable. As of February 28, 2021, we wholly-owned a pending U.S. provisional patent application relating to RLY-4008 composition of matter, methods of treatment, solid forms and methods of manufacture. Any U.S. or foreign patent that may issue from a non-provisional patent application claiming priority

to this application would be scheduled to expire in 2041, excluding any additional term for patent term adjustment or patent term extension, if applicable.

Mutant-PI3Kα Inhibitor Program

As of February 28, 2021, we co-owned with D.E. Shaw Research pending U.S. provisional patent applications that cover our PI3K program, which are directed to the composition of matter for the drug candidates of the program, analogs thereof, as well as methods of making and using these compounds. Any U.S. or foreign patent that may issue from a non-provisional patent application claiming priority to this applications would be scheduled to expire in 2041, excluding any additional term for patent term adjustment or patent term extension, if applicable.

RLY-1971

We wholly own a U.S. patent which relates to RLY-1971 composition of matter, which is scheduled to expire in 2039, excluding any additional term for patent term adjustment or patent term extension, if applicable. As of February 28, 2021, we co-owned with D. E. Shaw Research pending United States and foreign applications, which relate to SHP2 inhibitor compositions of matter and methods of treatment. Any U.S. or foreign patent that may issue from these patent applications would be scheduled to expire in 2039, excluding any additional term for patent term adjustment or patent term extension, if applicable. As of February 28, 2021, we wholly-owned pending U.S., PCT, Argentine and Taiwanese patent applications which relate to RLY-1971, solid forms and methods of manufacture. Any U.S. or foreign patent that may issue from a non-provisional patent application claiming priority to these patent applications would be scheduled to expire in 2040, excluding any additional term for patent term adjustment or patent term extension, if applicable.

Pursuant to the Genentech Agreement, we have granted an exclusive, worldwide, royalty-bearing license to Genentech, with the right to sublicense, develop and commercialize RLY-1971 and any other SHP2 inhibitors developed under the Genentech Agreement. Genentech has the first right, but not the obligation, to file, prosecute and maintain any patents licensed to it, as well as to enforce infringement of or defend claims against such patents that relate to RLY-1971 or other SHP2 inhibitors. See “—Our Collaborations—License Agreements and Strategic Collaborations—Genentech Collaboration and License Agreement” for more information on the Genentech Agreement.

Prosecution of the PCT patent application covering our FGFR2 inhibitors and the provisional patent application covering our PI3K program has not commenced, and will not commence unless and until they are timely converted into U.S. non-provisional or national stage applications. Prosecution is a lengthy process, during which the scope of the claims initially submitted for examination by the USPTO or other foreign jurisdiction are often significantly narrowed by the time they issue, if they issue at all. Any U.S. or foreign patent issuing from these provisional, PCT, or foreign patent applications (assuming they are timely converted into non-provisional applications, and such non-provisional applications are granted as issued patents) would be scheduled to expire in 2040 or 2041 (for our FGFR2 applications) or 2041 (for our PI3K application), excluding any additional term for patent term adjustment or patent term extension, and assuming national phase entries are timely made based upon the pending PCT application, and payment of all applicable maintenance or annuity fees. Any of our pending PCT patent applications are not eligible to become issued patents until, among other things, we file national stage patent applications within 30 months in the countries in which we seek patent protection. If we do not timely file any national stage patent applications, we may lose our priority date with respect to our PCT patent applications and any patent protection on the inventions disclosed in such PCT patent applications. Our provisional patent applications may never result in issued patents and are not eligible to become issued patents until, among other things, we file a non-provisional patent application and/or PCT patent application within 12 months of filing the related provisional patent application. If we do not timely file non-provisional patent applications, we may lose our priority date with respect to our provisional patent applications and any patent protection on the inventions disclosed in our provisional patent applications. While we intend to timely file non-provisional and national stage patent applications relating to our provisional and PCT patent applications, we cannot predict whether any of our current or future patent applications for any of our product candidates or technology, will issue as patents. If we do not successfully obtain patent protection, or, even if we do obtain patent protection, if the scope of the patent protection we, Genentech, or our potential licensors, obtain with respect to any of our product candidates or technology is not sufficiently broad, we will be unable to prevent others

from using our technology or from developing or commercializing technology and products similar or identical to ours or other competing products and technologies.

In addition to patent applications, we rely on unpatented trade secrets, know-how and continuing technological innovation to develop and maintain our competitive position. However, trade secrets and confidential know-how are difficult to protect. In particular, we anticipate that with respect to the building of our compound library, our trade secrets and know-how will over time be disseminated within the industry through independent development and public presentations describing the methodology. We seek to protect our proprietary information, in part, by executing confidentiality agreements with our collaborators and scientific advisors and non-competition, non-solicitation, confidentiality and invention assignment agreements with our employees and consultants. We have also executed agreements requiring assignment of inventions with selected consultants, scientific advisors and collaborators. The confidentiality agreements we enter into are designed to protect our proprietary information and the agreements or clauses requiring assignment of inventions to us are designed to grant us ownership of technologies that are developed through our relationship with the respective counterparty. We cannot guarantee that we will have executed such agreements with all applicable employees and contractors, or that these agreements will afford us adequate protection of our intellectual property and proprietary information rights. In addition, our trade secrets and/or confidential know-how may become known or be independently developed by a third party or misused by any collaborator to whom we disclose such information. These agreements may also be breached, and we may not have an adequate remedy for any such breach. Despite any measures taken to protect our intellectual property, unauthorized parties may attempt to copy aspects of our products or to obtain or use information that we regard as proprietary. Although we take steps to protect our proprietary information, third parties may independently develop the same or similar proprietary information or may otherwise gain access to our proprietary information. As a result, we may be unable to meaningfully protect our trade secrets and proprietary information. For more information regarding the risks related to our intellectual property, please see “Risk Factors—Risks Related to our Intellectual Property.”

Commercialization

Subject to receiving marketing approvals, we expect to commence commercialization activities by building a focused sales and marketing organization in the United States to sell our products. We believe that such an organization will be able to address the community of oncologists who are the key specialists in treating the patient populations for which our product candidates are being developed. Outside the United States, we expect to enter into distribution and other marketing arrangements with third parties for any of our product candidates that obtain marketing approval.

We also plan to build a marketing and sales management organization to create and implement marketing strategies for any products that we market through our own sales organization and to oversee and support our sales force. The responsibilities of the marketing organization would include developing educational initiatives with respect to approved products and establishing relationships with researchers and practitioners in relevant fields of medicine.

Manufacturing

We do not have any manufacturing facilities or personnel. We currently rely, and expect to continue to rely, on third parties for the manufacture of our product candidates undergoing preclinical testing, as well as for clinical testing and commercial manufacture if our product candidates receive marketing approval.

All of our drug candidates are small molecules and are manufactured in synthetic processes from available starting materials. The chemistry appears amenable to scale-up and does not currently require unusual equipment in the manufacturing process. We expect to continue to develop product candidates that can be produced cost-effectively at contract manufacturing facilities.

We generally expect to rely on third parties for the manufacture of companion diagnostics for our products, which are assays or tests to identify an appropriate patient population. Depending on the technology solutions we choose, we may rely on multiple third parties to manufacture and sell a single test.

Governmental Regulation

The FDA and other regulatory authorities at federal, state and local levels, as well as in foreign countries, extensively regulate, among other things, the research, development, testing, manufacture, quality control, import, export, safety, effectiveness, labeling, packaging, storage, distribution, recordkeeping, approval, advertising, promotion, marketing, post-approval monitoring and post-approval reporting of drugs. We, along with our vendors, contract research organizations and contract manufacturers, will be required to navigate the various preclinical, clinical, manufacturing and commercial approval requirements of the governing regulatory agencies of the countries in which we wish to conduct studies or seek approval of our product candidates. The process of obtaining regulatory approvals of drugs and ensuring subsequent compliance with appropriate federal, state, local and foreign statutes and regulations requires the expenditure of substantial time and financial resources.

In the United States, where we are initially focusing our drug development, the FDA regulates drug products under the Federal Food, Drug, and Cosmetic Act, or FD&C Act, as amended, its implementing regulations and other laws. If we fail to comply with applicable FDA or other requirements at any time with respect to product development, clinical testing, approval or any other legal requirements relating to product manufacture, processing, handling, storage, quality control, safety, marketing, advertising, promotion, packaging, labeling, export, import, distribution, or sale, we may become subject to administrative or judicial sanctions or other legal consequences. These sanctions or consequences could include, among other things, the FDA’s refusal to approve pending applications, issuance of clinical holds for ongoing studies, suspension or revocation of approved applications, warning or untitled letters, product withdrawals or recalls, product seizures, relabeling or repackaging, total or partial suspensions of manufacturing or distribution, injunctions, fines, civil penalties or criminal prosecution.

The process required by the FDA before our product candidates are approved as drugs for therapeutic indications and may be marketed in the United States generally involves the following:

• submission to the FDA of a New Drug Application, or NDA;

• payment of user fees for FDA review of the NDA; and

Preclinical studies and clinical trials for drugs

Before testing any drug in humans, the product candidate must undergo rigorous preclinical testing. Preclinical studies include laboratory evaluations of drug chemistry, formulation and stability, as well as in vitro and animal studies to assess safety and in some cases to establish the rationale for therapeutic use. The conduct of preclinical studies is subject to federal and state regulation, including GLP requirements for safety/toxicology studies. The results of the preclinical studies, together with manufacturing information and analytical data, must be submitted to the FDA as part of an IND. An IND is a request for authorization from the FDA to administer an investigational product to humans and must become effective before clinical trials may begin. Some long-term preclinical testing may continue after the IND is submitted. The IND automatically becomes effective 30 days after receipt by the FDA, unless the FDA, within the 30-day time period, raises concerns or questions about the conduct of the clinical trial, including concerns that human research subjects will be exposed to unreasonable health risks, and imposes a full or partial clinical hold. FDA must notify the sponsor of the grounds for the hold and any identified deficiencies must be resolved before the clinical trial can begin. Submission of an IND may result in the FDA not allowing clinical trials to commence or not allowing clinical trials to commence on the terms originally specified in the IND. A clinical hold can also be imposed once a trial has already begun, thereby halting the trial until the deficiencies articulated by FDA are corrected.

The clinical stage of development involves the administration of the product candidate to healthy volunteers or patients under the supervision of qualified investigators, who generally are physicians not employed by or under the trial sponsor’s control, in accordance with GCP requirements, which include the requirements that all research subjects provide their informed consent for their participation in any clinical trial. Clinical trials are conducted under protocols detailing, among other things, the objectives of the clinical trial, dosing procedures, subject selection and exclusion criteria and the parameters and criteria to be used in monitoring safety and evaluating effectiveness. Each protocol, and any subsequent amendments to the protocol, must be submitted to the FDA as part of the IND. Furthermore, each clinical trial must be reviewed and approved by an IRB for each institution at which the clinical trial will be conducted to ensure that the risks to individuals participating in the clinical trials are minimized and are reasonable compared to the anticipated benefits. The IRB also approves the informed consent form that must be provided to each clinical trial subject or his or her legal representative and must monitor the clinical trial until completed. The FDA, the IRB, or the sponsor may suspend or discontinue a clinical trial at any time on various grounds, including a finding that the subjects are being exposed to an unacceptable health risk. There also are requirements governing the reporting of ongoing clinical trials and completed clinical trials to public registries. Information about clinical trials, including results for clinical trials other than Phase 1 investigations, must be submitted within specific timeframes for publication on www.ClinicalTrials.gov, a clinical trials database maintained by the National Institutes of Health.

A sponsor who wishes to conduct a clinical trial outside of the United States may, but need not, obtain FDA authorization to conduct the clinical trial under an IND. If a foreign clinical trial is not conducted under an IND, FDA will nevertheless accept the results of the study in support of an NDA if the study was conducted in accordance with GCP requirements, and the FDA is able to validate the data through an onsite inspection if deemed necessary.

Clinical trials to evaluate therapeutic indications to support NDAs for marketing approval are typically conducted in three sequential phases, which may overlap.

In August 2018, the FDA released a draft guidance entitled “Expansion Cohorts: Use in First-In-Human Clinical Trials to Expedite Development of Oncology Drugs and Biologics,” which outlines how drug developers can utilize an adaptive trial design commonly referred to as a seamless trial design in early stages of oncology drug development (i.e., the first-in-human clinical trial) to compress the traditional three phases of trials into one continuous trial called an expansion cohort trial. Information to support the design of individual expansion cohorts are included in IND applications and assessed by FDA. Expansion cohort trials can potentially bring efficiency to drug development and reduce development costs and time.

Post-approval trials, sometimes referred to as Phase 4 clinical trials or post-marketing studies, may be conducted after initial marketing approval. These trials are used to gain additional experience from the treatment of patients in the intended therapeutic indication and are commonly intended to generate additional safety data regarding use of the product in a clinical setting. In certain instances, the FDA may mandate the performance of Phase 4 clinical trials as a condition of NDA approval.

RLAY US Equity

RLAY · 10-K · period ended 2020-12-31

Item 1. Business.

RLAY US Equity☆

RLAY · 10-K · period ended 2020-12-31

Item 1. Business.

RLAY US Equity