Rzepiela, A. | Search bioRxiv

bioRxiv2026

Point-Cloud Enhancement and Structural Interpretation Framework for Cryo-EM Maps

Cryo-Electron Microscopy (cryo-EM) has become a cornerstone of modern structural biochemistry, enabling the reconstruction of 3D protein maps at near-atomic resolution. Despite its transformative impact, interpreting maps remains challenging. Structural heterogeneity and molecular flexibility often produce low-resolution densities, which post-processing can only sharpen in well-ordered regions. This leaves flexible areas either poorly resolved or lost. Recent deep-learning approaches have demonstrated strong potential for enhancing cryo-EM maps, enabling an extended interpretation of cryo-EM densities. Most of these methods operate on small volumetric blocks, which restricts the receptive field of the model and prevents it from leveraging the broader structural context of the protein. To address this limitation, we introduce CryoPC, a model that enriches local block-based processing with a compact point-cloud representation of the entire map. By conditioning the network on this global representation, CryoPC is able to capture both local and distant structural features, allowing it to make more globally consistent predictions. We demonstrate that incorporating this global context consistently improves enhancement quality across a variety of samples. Furthermore, CryoPC achieves great performance compared to existing methods of similar speed, offering a practical and scalable solution for cryo-EM map enhancement. Finally, we present a framework for assesing a quality of enhanced maps using metrics from the Phenix package.

bioinformatics↗

bioRxiv2023

Refinement of Cryo-EM 3D Maps with Self-Supervised Denoising Model: crefDenoiser

Cryogenic electron microscopy (cryo-EM) is a pivotal technique for imaging macromolecular structures. Despite extensive processing of large image sets collected in a cryo-EM experiment to amplify the signal-to-noise ratio, the reconstructed 3D protein density maps are often limited in quality due to residual noise, which in turn affects the accuracy of the macromolecular representation. In this paper, we introduce crefDenoiser, a denoising neural network model designed to enhance the signal in 3D cryo-EM maps produced with standard processing pipelines, beyond the current state of the art. crefDenoiser is trained without the need for clean, ground-truth target maps. Instead, we employ a custom dataset composed of real noisy protein half-maps sourced from the Electron Microscopy Data Bank repository. Strong model performance is achieved by optimizing for the theoretical noise-free map during self-supervised training. We demonstrate that our model successfully amplifies the signal across a wide variety of protein maps, outperforming a classical map denoiser and a network-based sharpening model. Without biasing the map, the proposed denoising method often leads to improved visibility of protein structural features, including protein domains, secondary structure elements, and amino-acid side chains.

bioinformatics↗