Protein interface design improved by joint sequence and structure diffusion
Joint Discrete and SE(3) Diffusion for Protein Interface Design
Artificial Intelligence
Summary
Designing proteins that bind tightly and specifically to other proteins is important for new medicines but is hard because shapes and chemistry must both fit well. The authors present PIDiff, a method that creates protein sequences and shapes together using a special type of mathematical diffusion that handles both discrete sequences and 3D structures. They built PIDiff’s core with PIFormer, which pays attention to protein geometry and sequence simultaneously. They trained it on a new diverse protein dataset called PIBench and showed it designs better protein binders for therapy-related targets than previous methods.
What this means in practice
- •For drug developers: Create custom protein binders that target disease proteins more accurately by designing both sequence and shape jointly.$Commercial implications: Enables companies to develop novel therapeutic proteins with higher binding quality and specificity than current design tools.
- •For biotech tool builders: Integrate a dual diffusion model approach to improve automated protein interface design in software platforms.
Authors
Zhenqiao Song, Jianwen Xie, Lei Li
Abstract
The design of high-affinity, target-specific protein binders is central to the development of modern therapeutics. However, existing protein-generation methods remain limited in this setting, as many generated designs fail to simultaneously achieve geometric and physicochemical complementarity at binding interfaces. To address this challenge, we introduce PIDiff, a physicochemically enhanced multimodal diffusion framework for protein-protein complex design. PIDiff jointly designs target-conditioned protein sequences and structures by coupling diffusion in discrete sequence space with diffusion on the continuous (SE(3)) structural manifold. At its core, PIDiff employs PIFormer, a novel architecture that interleaves geometry-aware invariant point attention with global sequence self-attention. We further curate PIBench, a protein-protein complex dataset with diverse target scaffolds derived from the Protein Data Bank (PDB). We first pretrain PIDiff on PIBench and subsequently fine-tune it for two therapeutically relevant applications: minibinder design and antibody design. PPFrame consistently outperforms baseline methods across sequence design success rate, structure binding energy, and sequence-structure co-designability on all tasks.