Oof.

My least favorite type of investor is the cheerleader. A cheerleader is someone who never admits they're wrong, is incapable of being objective, and makes emotional decisions with money. They usually take a top-down approach, which dresses up regurgitated and unoriginal summaries of basic information as analysis and research. There are a lot of cheerleaders out there.

Objectively speaking, the initial safety and tolerability data from RLY-2608 leave no room for doubt about the value of utilizing protein motion to design more selective drug compounds. Relay Therapeutics has crushed that part of its hypothesis for the Dynamo technology platform across the first three assets with clinical data.

Objectively speaking, the initial efficacy signals for RLY-2608 were not robust. Although the first part of the phase 1 study was not intended to evaluate efficacy, the surprisingly lackluster early results suggest the core hypothesis for the asset – a more tolerable pan-mutant inhibitor can lead to improved outcomes – might be at risk.

A careful analysis of the preliminary phase 1 data reveals nuances for PI3K-alpha, RLY-2608, and the competitive landscape. It's frustrating, but these nuances lead to questions that cannot be answered with existing data.

Let's dig in, specifically focusing on the breast cancer cohorts.

Safety Data: Another No-Doubter for Dynamo

PI3K-alpha (also called PIK3CA) is the second-most common driver of solid tumor cancers that remains inadequately drugged. The protein is mutated in an estimated 14% of all solid tumor cancers, which means the global revenue opportunity up for grabs likely exceeds $15 billion. The only mutation that's more common is KRAS, harbored by an estimated 22% of solid tumors. In fact, PI3K-alpha and KRAS mutations often coexist (more on that below #hinthint).

But it's a delicate target. PI3K-alpha regulates glucose metabolism, so drug compounds that inhibit both mutated (in tumors) and wildtype (healthy working copies) versions of the protein cause unsafe increases in glucose levels. This is called hyperglycemia. As a result, patients with diabetes cannot be treated with the only FDA-approved PI3K-alpha inhibitor, Piqray (alpelisib) from Novartis.

There are also multiple structures, or isoforms, of the PI3K protein. Drug compounds that whack the delta isoform (PI3K-delta) cause gastrointestinal side effects such as diarrhea. The beta isoform shouldn't be trampled either.

Hyperglycemia, diarrhea, and rash are the three most common dose-limiting toxicities (DLTs) of Piqray and other development-stage inhibitors in this class that aren't mutant or isoform selective. Relay Therapeutics designed RLY-2608 to be both mutant and isoform specific, which is clear from the initial safety data.

Consider the safety and tolerability profiles of Piqray (using phase 3 data), inavolisib (phase 1b), and RLY-2608 (preliminary phase 1a) when combined with the chemotherapy drug fulvestrant. Grade 1-2 adverse events are moderate but manageable, while Grade 3-4 adverse events are more severe and can lead to treatment discontinuation:

‍

‍

A few notes:

• A drug product label must list adverse events that occur in at least 20% of all patients from a pivotal study. Similar to RLY-4008 (FGFR2), Relay Therapeutics is positioned to have the cleanest safety label in the class. Adverse events are likely to increase as the company explores higher doses.
• The data for RLY-2608 are from the dose escalation part of the phase 1 clinical trial. In other words, the primary goal here was to explore the safety of various dose levels (from 100 mg to 800 mg) and capture early efficacy signals. The ReDiscover study tested seven different dose levels including once-daily and twice-daily dosing.
• The upcoming dose expansion cohort of the phase 1 clinical trial will enroll more patients and test fewer dosing levels. The company signaled it will likely settle on twice-daily, mid- to high-level dosing and may explore dose levels above 800 mg.

Due to DLTs, Piqray combined with fulvestrant (phase 3) only achieved a median duration of exposure (the duration of treatment excluding periods of dose interruptions) of 22 weeks and a median dose intensity (the percentage of patients receiving the prescribed dose level across the study period) of 82.7%. These are the maximum levels achievable for the asset.

Inavolisib combined with fulvestrant (phase 1b) achieved a median duration of treatment of 19 weeks and a median dose intensity of 98%. Genentech reported that 60% of individuals discontinued treatment due to disease progression. The data in the table above are from a single dose level.

By contrast, RLY-2608 combined with fulvestrant (preliminary phase 1a) achieved a median duration of treatment of 16 weeks and counting (70% of patients remained on treatment at the data cutoff). The median dose intensity was greater than 98%. The maximum levels have not been achieved and dose levels are still being optimized.

Despite discontinuous treatment requirements to navigate tolerability issues, Piqray generated $400 million in full-year 2022 revenue for Novartis. It earned FDA approval in mid-2019.

Efficacy Data: Not So Great, But Early

Although RLY-2608 avoided DLTs and achieved a high dose intensity, the asset appeared to have a curiously low response rate through the data cutoff (solely for inclusion in the medical conference) in the dose escalation part of the phase 1 study. That's why the stock plunged 36%.

I was initially downbeat in my comments made after only reviewing the limited information in the press release. However, an objective analysis of the data from ReDiscover and the competitive landscape shows the asset is on a similar trajectory to both Piqray and inavolisib. The safety profile still positions RLY-2608 to best both competitors.

The reason PI3K-alpha is an important target for treating cancer, aside from the large, underserved patient populations, is due to where it sits within cellular growth pathways. It's downstream of KRAS, but upstream of other important cellular signaling proteins such as PIP2, PIP3, PTEN, FOXO, mTOR, AKT, and others – each a large opportunity. The unique mechanism of activation of mutant PI3K proteins and ability of tumor cells to reroute communication signals requires long duration of inhibition to drive a response. Molecules that aren't mutant or isoform selective lead to too many DLTs, which limits their ability to deliver long duration of inhibition and their efficacy.

This was clearly observed in studies of Piqray and inavolisib, where median metrics hid the extremes.

• Individuals who didn't respond generally experienced frequent dose interruptions had short duration of treatment and exposure.
• Individuals who had partial responses generally experienced fewer dose interruptions and longer duration of treatment.

Consider the phase 1b trial of inavolisib combined with fulvestrant.

• Genentech reported an objective response rate (ORR) of 26% (14/54) in individuals with measurable disease.
• The median duration of treatment for all patients was 22 weeks.
• The median duration of treatment of a partial response (PR) was 34 weeks.
• The median duration of treatment of progressive disease (PD) was only 8 weeks.
• The study measured the best response achieved by each patient. Acknowledging that nuance, half of patients who achieved a PR discontinued treatment by the end of the study, suggesting responses weren't durable.

‍

As the first PI3K-alpha inhibitor, studies of Piqray were designed differently. The phase 1 study evaluated Piqray as a monotherapy. Subsequent studies combined with fulvestrant focused on progression free survival (PFS) and didn't publish response data in similar formats. Nonetheless, the phase 3 study had a median follow up of 80 weeks, but the median duration of exposure was only 22 weeks. The gap is explained by the extremes between responders (who drove the ORR) and progressors (who tanked the duration of exposure and treatment).

Now consider the initial efficacy signals from the dose escalation part of the ReDiscover study, which wasn't designed to provide interpretable efficacy data. Keep in mind this study tested numerous dose levels. Focusing on the breast cancer cohorts with measurable disease:

• The median duration of treatment for all patients was 16 weeks and counting (only due to the data cutoff for presenting at the medical conference).
• The median duration of treatment for patients with confirmed tumor reductions was 20 weeks. Of these nine patients, 89% (8/9) remained on treatment at the data cutoff. One had a PR confirmed after the data cutoff.
• The median duration of treatment for patients with PD was only 8 weeks.

‍

The data for RLY-2608 and comparisons within the competitive landscape get even trickier when additional nuances are considered. Thanks to nerds in lab coats, we need to define a handful of terms:

• RLY-2608 is a pan-mutant inhibitor, which means the molecule has activity across mutations. Think panorama, pandemonium, or Pantera. m/_(> . <)_m/. Specifically, the molecule targets three groups of mutations on PI3K-alpha. The ReDiscover study enrolled patients with 13 unique mutation combinations.
• The data discussed above for Piqray, inavolisib, and RLY-2608 only include patients with measurable disease, which means individuals with breast tumors that can be measured radiologically. Individuals with non-measurable disease also have bone and central nervous system (CNS) lesions that are less well defined. These patients are typically separated and have different rates of response. For example, a second individual in the ReDiscover study achieved a breast tumor reduction of at least 30%, which would qualify as a PR. However, the emergence of a bone lesion means this individual counts as stable disease (SD) according to clinical criteria. If counted, then RLY-2608 would have an ORR of 12.5% (2/16) despite various dose levels and an arbitrary data cutoff. That's a very promising trajectory for a PI3K-alpha inhibitor after only 16 weeks of treatment.
• Individuals with measurable disease are categorized as PD, SD, PR, and complete response (CR). Individuals with non-measurable disease are categorized as PD, non-CR / non-PD, and PD.
• Studies of Piqray, inavolisib, and RLY-2608 measured the circulating tumor DNA (ctDNA) shed by tumors to gauge the inhibition of PI3K-alpha. RLY-2608 significantly outpaced the competitive landscape despite having the least mature data, suggesting responses will further improve with time. This was also observed from RLY-4008 (FGFR2). The asset had an initial ORR of 50%, which improved to 88% once more complete phase 1 data were available with an optimized dose level. The ability to avoid DLTs allowed a longer duration of treatment, which drove responses over time – a capability unique to Dynamo-designed molecules.

Uncertainty Remains

To be clear, RLY-2608 isn't a slam dunk. But the data aren't nearly as poor as the stock movement suggests.

By designing a truly selective pan-mutant inhibitor, Relay Therapeutics might prove that targeting PI3K-alpha alone is inadequate. Humanity hasn't been able to test this hypothesis before due to dose-limiting toxicities. There are nerdy cellular communication pathway details to consider. Breast cancers and head and neck cancers can be heavily influenced by PI3K-alpha mutations. However, there are data suggesting PI3K-alpha mutations are a weak driver of other solid tumors. That could significantly reduce the expected market opportunity for RLY-2608 (or RLY-5836), although it could still generate billions of dollars in peak sales in the U.S. from breast cancer populations alone.

Patient selection could be one reason for the seemingly lackluster early efficacy signals from ReDiscover. One study observed that Piqray provided no benefit when tumors also harbored mutations in KRAS, TP53, or FGFR1. That suggests new combination treatments might be required to shrink tumors in these patient populations, which would be more feasible for PI3K-alpha inhibitors with favorable tolerability profiles. It's unclear if Relay Therapeutics sequenced tumors (or ctDNA) for concurrent mutations in ReDiscover. Concurrent mutations are not part of the exclusion criteria for the phase 1 study.

The safety and tolerability of RLY-2608 suggests it has significant combination potential with other agents, including KRAS, FGFR1, CDK2, and other inhibitors or protein degraders. The data for Piqray, inavolisib, and RLY-2608 presented above are specifically for HR+, HER2- breast cancers and combinations with fulvestrant. Fulvestrant is a chemotherapy that inhibits estrogen receptor (ER) proteins – meaning it takes care of the HR+ part of these tumors. But treating a broader patient population, including within breast cancers, may require triple combinations. In other words, it's still early, but limiting DLTs provides significant optionality. Relay Therapeutics is building a breast cancer pipeline around PI3K-alpha inhibitors for this reason, including FGFR2, CDK2, ER-alpha, and other inhibitors and degraders.

The dose expansion part of the phase 1 study includes a triple combination arm.

Forecast & Modeling Insights

(No change.)

This is new territory for modelers like the finch.

On the one hand, RLY-2608 completely avoided treatment discontinuations from adverse events. That creates the potential to evaluate combinations with other agents (not only with fulvestrant), observe responses over much longer time periods, and/or evaluate higher dose levels. Additionally, investigators can confirm the molecule pummeled PI3K-alpha mutations.

On the other hand, initial efficacy data are surprisingly lackluster despite these advantages. It's simply too early to provide definitive conclusions about the efficacy of RLY-2608. Solt DB Invest's analysis suggests there's significant potential for more mature data to deliver favorable outcomes, although there are metabolic uncertainties that could tank the company's overall hypothesis for combinations with fulvestrant alone.

The stock tumbled 36% due to the increased uncertainty for the future of the asset, but investors cannot make definitive conclusions about the efficacy of '2608 until more complete data – more individuals, more focused dose levels, and longer treatment duration – become available in 2024.

One update to our forecast and modeling:

• Solt DB Invest no longer expects RLY-5836 to definitively replace RLY-2608 as the company's pan-mutant PI3K-alpha inhibitor. The fast follower asset was meant to be a hedge against potential endocrine disruption caused by '2608, although no endocrine disruption was observed in the dose escalation part of the phase 1 clinical trial. RLY-5836 began a phase 1 clinical trial in April 2023 and is expected to complete the study in 2025. Relay Therapeutics may advance one or both assets into mid-stage trials.

Margin of Safety & Allocation

(No change.)

Relay Therapeutics is considered a Growth (Quality) position. The current fair valuation for the company based on our 2023 model is below:

• Current Price (market close April 19): $12.35 per share‍
• Modeled Fair Valuation:   $26.21 per share‍
• Allocation Range:             Up to 15%

Relay Therapeutics reported 121.384 million shares outstanding as of February 17, 2023. The modeled fair valuation above assumes 133.523 million shares outstanding.

Further Reading

• April 2023 press release announcing preliminary phase 1a data for RLY-2608
• April 2023 research note discussing our updated model for Relay Therapeutics due to higher estimated probability of success (POS) outcomes
• Phase 1 results for Piqray and fulvestrant
• Phase 3 results for Piqray and fulvestrant combination
• Phase 1b results for inavolisib and fulvestrant

‍