SaDiT: Efficient Protein Backbone Design via Latent Structural Tokenization and Diffusion Transformers

2026-02-06

AI Summary

This work addresses the high computational cost and low sampling efficiency of existing diffusion-based protein backbone generation methods, which hinder large-scale structural exploration. The authors propose SaDiT, a novel framework that, for the first time, integrates SaProt’s structural tokenization with a Diffusion Transformer (DiT) to model protein geometry in a discrete latent space while preserving SE(3) equivariance. By introducing an IPA Token Cache mechanism to optimize attention computation, SaDiT significantly accelerates generation speed. In both unconditional and fold-class-conditional generation tasks, SaDiT outperforms state-of-the-art models such as RFDiffusion and Proteina in terms of computational efficiency and structural designability.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Natural Language Processing: Generation
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Large language models for search
• Semantics and Knowledge: Scalable techniques for the creation, curation, publication, maintenance, and consumption of large, Web-based, structured, reusable, knowledge graphs and ontologies

Abstract

Generative models for de novo protein backbone design have achieved remarkable success in creating novel protein structures. However, these diffusion-based approaches remain computationally intensive and slower than desired for large-scale structural exploration. While recent efforts like Proteina have introduced flow-matching to improve sampling efficiency, the potential of tokenization for structural compression and acceleration remains largely unexplored in the protein domain. In this work, we present SaDiT, a novel framework that accelerates protein backbone generation by integrating SaProt Tokenization with a Diffusion Transformer (DiT) architecture. SaDiT leverages a discrete latent space to represent protein geometry, significantly reducing the complexity of the generation process while maintaining theoretical SE(3) equivalence. To further enhance efficiency, we introduce an IPA Token Cache mechanism that optimizes the Invariant Point Attention (IPA) layers by reusing computed token states during iterative sampling. Experimental results demonstrate that SaDiT outperforms state-of-the-art models, including RFDiffusion and Proteina, in both computational speed and structural viability. We evaluate our model across unconditional backbone generation and fold-class conditional generation tasks, where SaDiT shows superior ability to capture complex topological features with high designability.

Problem

• protein backbone design
• computational efficiency
• structural tokenization
• diffusion models
• generative modeling

Innovation

• Latent Structural Tokenization
• Diffusion Transformer
• IPA Token Cache
• Protein Backbone Design
• SE(3) Equivariance