DUAL: deep unsupervised simultaneous simulation and denoising for cryo-electron tomography

Abstract

Recent biotechnological developments in cryo-electron tomography allow direct visualization of native sub-cellular structures with unprecedented details and provide essential information on protein functions/dysfunctions. Denoising can enhance the visualization of protein structures and distributions. Automatic annotation via data simulation can ameliorate the time-consuming manual labeling of large-scale datasets. Here, we combine the two major cryo-ET tasks together in DUAL, by a specific cyclic generative adversarial network with novel noise disentanglement. This enables end-to-end unsupervised learning that requires no labeled data for training. The denoising branch outperforms existing works and substantially improves downstream particle picking accuracy on benchmark datasets. The simulation branch provides learning-based cryo-ET simulation for the first time and generates synthetic tomograms indistinguishable from experimental ones. Through comprehensive evaluations, we showcase the effectiveness of DUAL in detecting macromolecular complexes across a wide range of molecular weights in experimental datasets. The versatility of DUAL is expected to empower cryo-ET researchers by improving visual interpretability, enhancing structural detection accuracy, expediting annotation processes, facilitating cross-domain model adaptability, and compensating for missing wedge artifacts. Our work represents a significant advancement in the unsupervised mining of protein structures in cryo-ET, offering a multifaceted tool that facilitates cryo-ET research.

CryoSamba: self-supervised deep volumetric denoising for cryo-electron tomography data

“…CryoSamba circumvents the inherent limitations common to denoising techniques that rely on synthetic data (Zeng et al, 2024), which may not generalize effectively to experimental images of varying characteristics or resolutions. It also avoids the pitfalls of methods that depend on noise modeling (Li et al, 2022), which can fail to fully grasp the complexities of noise in experimental data.…”

“…Recently, deep learning approaches have emerged as superior alternatives (Buchholz et al, 2019;Bepler et al, 2020). These methods adapt to the intricacies of the data, but because cryo-ET data generally lack ground truth high SNR images for direct supervision, most deep-learning denoising algorithms rely on self-supervision (Lehtinen et al, 2018), using paired 3D volumes from an even/odd split of the cryo-ET tilt-series (Buchholz et al, 2019;Bepler et al, 2020), synthetic or annotated data (Zeng et al, 2024), or noise modeling (Li et al, 2022).…”