CryoET-DDM: restoring high-fidelity signals in cryo-electron tomography via a data-driven diffusion model
Bioinformatics · 12.11.2026
DOI: 10.1093/bioinformatics/btag824 ↗Kurz gefasst
Fachartikel in der Zeitschrift Bioinformatics, online erschienen am 12.11.2026. Verfasst von Xinsheng Wang und 7 weiteren Autor:innen. Beteiligte Einrichtung: Key Laboratory of Brain Health Intelligent Evaluation and Intervention, Ministry of Education, Beijing Institute of Technology, 100081, Beijing, Chin…. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Motivation Cryo-electron tomography (cryo-ET) is an essential technique for resolving high-resolution structures of macromolecular complexes in situ. However, the extremely low signal-to-noise ratio (SNR) in tilt series, caused by strict electron dose limitations, remains a major bottleneck for subtomogram averaging (STA). Although existing denoising methods can enhance tomogram contrast, they may compromise the high-frequency structural information required for downstream STA refinement. Results We introduce CryoET-DDM, a diffusion framework for high-fidelity restoration of cryo-ET tilt series that integrates data-driven noise modeling and fidelity-guided structural preservation. CryoET-DDM incorporates background noise patches extracted from the same experimental tilt series into the diffusion process to model the noise characteristics of low-dose imaging. A fidelity-guided learning objective promotes structural consistency during denoising, helping preserve high-frequency information needed for STA. Experiments on three public datasets (EMPIAR-10499, EMPIAR-10164, and EMPIAR-10651) show that CryoET-DDM improves particle quality, preserves structural fidelity, and yields downstream gains in STA resolution. These results support tilt series denoising as a promising strategy for cryo-ET data processing and improve the reliability of in situ structural biology analysis. Availability and implementation Code and models for our experiments are available at https://github.com/xinsheng44/CryoET-DDM. The real datasets used in this study are publicly available from EMPIAR.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Key Laboratory of Brain Health Intelligent Evaluation and Intervention, Ministry of Education, Beijing Institute of Technology, 100081, Beijing, China.
- School of Medical Science and Engineering, Beijing Institute of Technology, 100081, Beijing, China.
Zitieren
APA
Wang, X., Li, F., Zhang, C., Tang, S., Jia, Z., Hu, B., Zhang, F., & Wan, X. (2026). CryoET-DDM: restoring high-fidelity signals in cryo-electron tomography via a data-driven diffusion model. Bioinformatics. https://doi.org/10.1093/bioinformatics/btag824
Vancouver
Wang X, Li F, Zhang C, Tang S, Jia Z, Hu B, et al. CryoET-DDM: restoring high-fidelity signals in cryo-electron tomography via a data-driven diffusion model. Bioinformatics. 2026. doi:10.1093/bioinformatics/btag824
BibTeX
@article{Wang2026,
author = {Wang, Xinsheng and Li, Fuwei and Zhang, Chuanyang and Tang, Shuai and Jia, Zhuo and Hu, Bin and Zhang, Fa and Wan, Xiaohua},
title = {CryoET-DDM: restoring high-fidelity signals in cryo-electron tomography via a data-driven diffusion model},
journal = {Bioinformatics},
year = {2026},
doi = {10.1093/bioinformatics/btag824}
}
EUVIMED