Efficient and Accurate Zero Shot Generation of Symmetric Protein Complexes
En palabras de los autores
Symmetric organization is a fundamental principle of biomolecular complex assembly. The same principle underlies engineered applications, from vaccine platforms to nanomaterials, making symmetric de novo generation a core capability for computational protein design. We establish a theoretical foundation equating generation of an assembly with underlying symmetry group to generation within a quotient space over . This directly allows -equivariant models to generate symmetric structures with minimal distribution shift. With this insight we introduce ZeUS (ro-shot nified ymmetrization), a framework that converts any pretrained -equivariant backbone generative model into an efficient generator of symmetric assemblies with no retraining. ZeUS operates on a single and utilizes to preserve internal model operations, reducing memory and runtime cost by for finite groups , compared to a full-complex forward pass. We implement ZeUS for IPA-based models (FrameDiff, FoldFlow) and axial-attention-based RFdiffusion, showing high generation efficiency with minimal degradation to generation quality. We then apply ZeUS to generate symmetric complexes of over residues with exact, untruncated attention, a scale previously intractable with any model.
Apareció: martes, 10 de noviembre. arXiv. Preprint, todavía sin revisión por pares.
Comentario de los autores: Published as a conference paper at NeurIPS 2026