Relay Therapeutics

Traditional drug discovery struggles to identify actionable pockets within proteins implicated in cancer, especially for mutated enzymes like PIK3CA that cycle through multiple conformations. Existing small molecules could not selectively target the specific mutant conformations driving breast cancer progression.

Relay Therapeutics built a proprietary ML and computational platform trained on structural biology and biophysics data to simulate protein motion and identify novel binding pockets across a spectrum of protein conformations. The platform guided the design of RLY-2608, a small molecule targeting a specific PIK3CA mutant pocket that spends more time in an open state.

RLY-2608 combined with hormonal therapy reduced tumor size in approximately 81% of 31 study participants with measurable disease in a Phase 3 breast cancer trial, as announced in a June 2025 press release. The company successfully advanced from computational design to human Phase 3 testing, validating its AI-driven protein motion approach.

• AI-predicted protein conformational states can reveal druggable pockets missed by static structural biology methods.
• ML platforms do not replace medicinal chemists — chemist intuition remains critical for synthesizability screening of AI-generated hits.
• Proprietary training data on structural biology and biophysics is a key competitive moat for AI-driven biotech.