Only two cancer-driving genetic mutations, called oncogenes, are present in at least 10% of all cancers: KRAS and PI3K-alpha.

The PI3K-alpha protein plays a crucial role in cell growth, specifically orchestrating how glucose is shuttled into cells. That helps to explain why mutations in this gene are observed in an estimated 14% of all solid tumors – and why it's proven so difficult to safely inhibit the oncogene.

Relay Therapeutics used a combination of experimental and computational tools to better understand the nuances of the PI3K family of proteins.

First, it leveraged the Anton-2 supercomputer architecture to simulate 4D protein structure, or how 3D protein structure changes over time. The approach yielded insights into imperceptible differences between wild-type (healthy copies) and mutated PI3K-alpha proteins, as well as differences between isoforms (alpha, beta, delta, and gamma).

• For example, inhibiting wild-type PI3K-alpha interrupts glucose metabolism throughout the body, which causes a side effect known as hyperglycemia that mimics diabetes. An ideal PI3K-alpha inhibitor would be selective for mutated proteins but spare healthy copies of the protein outside of cancer cells.
• Meanwhile, drug compounds that broadly inhibit PI3K proteins can encounter tolerability issues. Inhibiting PI3K-delta proteins is associated with side effects such as diarrhea, vomiting, and suppression of bone marrow production. Whack PI3K-gamma and patients are more likely to develop a rash and mouth sores.

As a result, Relay Therapeutics was the first to solve the full-length structures of both wild-type and mutated PI3K-alpha proteins. It also discovered a novel allosteric pocket (a binding site distinct from the protein's primary active binding site) favored in certain mutated PI3K-alpha proteins, which provided it a novel target to design mutant-selective inhibitors against.

Second, it leveraged machine learning DNA-encoded libraries (ML-DEL or REL-DEL) to sift through billions of molecules in chemical space more efficiently. Relay Therapeutics searched for molecules expected to have basic pharmaceutical properties, binding activity against the newly discovered allosteric pocket, and limited binding activity against different PI3K isoforms (such as delta and gamma) and wild-type proteins.

The resulting drug candidate, RLY-2608, was the first mutant-selective, isoform-selective molecule to enter the clinic. The asset prefers binding to mutant proteins over wild-type proteins 5:1, while its potency against isoforms (beta >4,000:1, delta >9,000:1, and gamma >10,000:1) suggested off-target side effects could be limited.

Results from a preliminary phase 1 data readout in the first half of 2023 showed early promise. Patients receiving RLY-2608 developed hyperglycemia or diarrhea half as frequently as the industry's most-selective drug currently on the market. Relay Therapeutics expects the greatly improved tolerability of the asset to potentially make it the foundational building block for combination therapies.

More mature safety and efficacy, including from the first doublet and triplet combinations, are expected before the end of 2024.