How Helix Labs uses generative models to design proteins

Helix Labs sits at the frontier of AI-driven biology. Its pipeline begins with generative models that propose novel protein structures, then filters candidates through physics-based simulation and, finally, wet-lab validation.

This feature walks through how the company narrows billions of possibilities to a handful of testable molecules, the role of foundation models trained on biological data, and the partnerships with research hospitals that supply ground-truth feedback.

The result is a flywheel: better data sharpens the models, which propose better candidates.