Constrained Diffusion for Protein Design with Hard Structural Constraints

ICLR 2026 Conference Submission12985 Authors

Keywords: Constrained Diffusion, Generative Models, Protein Design, Proximal Optimization, Motif Scaffolding

TL;DR: This paper introduces a constrained diffusion framework for protein design that enforces functional, stereochemical, and geometric constraints via proximal updates, achieving state-of-the-art results on tasks like motif scaffolding and pocket design.

Abstract: Diffusion models offer a powerful means of capturing the manifold of realistic protein structures, enabling rapid design for protein engineering tasks. However, existing approaches observe critical failure modes when precise constraints are necessary for functional design. To this end, we present a constrained diffusion framework for structure-guided protein design, ensuring strict adherence to functional requirements while maintaining precise stereochemical and geometric feasibility. The approach integrates proximal feasibility updates with ADMM decomposition into the generative process, scaling effectively to the complex constraint sets of this domain. We evaluate on challenging protein design tasks, including motif scaffolding and vacancy-constrained pocket design, while introducing a novel curated benchmark dataset for motif scaffolding in the PDZ domain. Our approach achieves state-of-the-art, providing perfect satisfaction of bonding and geometric constraints with no degradation in structural diversity.

Primary Area: applications to physical sciences (physics, chemistry, biology, etc.)

Submission Number: 12985