RadDiff: Retrieval-Augmented Denoising Diffusion for Protein Inverse Folding

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Han, Jin, Fu, Tianfan, Li, Wu-Jun
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917323941609472
author Han, Jin
Fu, Tianfan
Li, Wu-Jun
author_facet Han, Jin
Fu, Tianfan
Li, Wu-Jun
contents Protein inverse folding, the design of an amino acid sequence based on a target protein structure, is a fundamental problem of computational protein engineering. Existing methods either generate sequences without leveraging external knowledge or relying on protein language models~(PLMs). The former omits the knowledge stored in natural protein data, while the latter is parameter-inefficient and inflexible to adapt to ever-growing protein data. To overcome the above drawbacks, in this paper we propose a novel method, called $\underline{\text{r}}$etrieval-$\underline{\text{a}}$ugmented $\underline{\text{d}}$enoising $\underline{\text{diff}}$usion~($\mbox{RadDiff}$), for protein inverse folding. In RadDiff, a novel retrieval-augmentation mechanism is designed to capture the up-to-date protein knowledge. We further design a knowledge-aware diffusion model that integrates this protein knowledge into the diffusion process via a lightweight module. Experimental results on the CATH, TS50, and PDB2022 datasets show that $\mbox{RadDiff}$ consistently outperforms existing methods, improving sequence recovery rate by up to 19\%. Experimental results also demonstrate that RadDiff generates highly foldable sequences and scales effectively with database size.
format Preprint
id arxiv_https___arxiv_org_abs_2512_00126
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle RadDiff: Retrieval-Augmented Denoising Diffusion for Protein Inverse Folding
Han, Jin
Fu, Tianfan
Li, Wu-Jun
Quantitative Methods
Artificial Intelligence
Protein inverse folding, the design of an amino acid sequence based on a target protein structure, is a fundamental problem of computational protein engineering. Existing methods either generate sequences without leveraging external knowledge or relying on protein language models~(PLMs). The former omits the knowledge stored in natural protein data, while the latter is parameter-inefficient and inflexible to adapt to ever-growing protein data. To overcome the above drawbacks, in this paper we propose a novel method, called $\underline{\text{r}}$etrieval-$\underline{\text{a}}$ugmented $\underline{\text{d}}$enoising $\underline{\text{diff}}$usion~($\mbox{RadDiff}$), for protein inverse folding. In RadDiff, a novel retrieval-augmentation mechanism is designed to capture the up-to-date protein knowledge. We further design a knowledge-aware diffusion model that integrates this protein knowledge into the diffusion process via a lightweight module. Experimental results on the CATH, TS50, and PDB2022 datasets show that $\mbox{RadDiff}$ consistently outperforms existing methods, improving sequence recovery rate by up to 19\%. Experimental results also demonstrate that RadDiff generates highly foldable sequences and scales effectively with database size.
title RadDiff: Retrieval-Augmented Denoising Diffusion for Protein Inverse Folding
topic Quantitative Methods
Artificial Intelligence
url https://arxiv.org/abs/2512.00126