Cryogenic electron microscopy (cryo-EM) is a widely used technique for determining the near-atomic resolution structures of biological macromolecules. In sample preparation, purified macromolecules are vitrified in a thin aqueous film. A transmission electron microscope equipped with a direct electron-detection camera is then used to capture two-dimensional projection images of individual particles. Due to random particle orientations within the ice layer relative to the electron beam, projections from various viewpoints are generated. Consequently, accurately determining the orientation angles of these two-dimensional projections constitutes a fundamental prerequisite for reconstructing the three-dimensional structure of the macromolecule. Ideally, particle orientations should be uniformly distributed; however, such uniformity is seldom achieved in practice—a consequence of the sample vitrification process. Owing to interfacial interactions, particles exhibit preferential adhesion to either the air-water or support-water interface of the thin film. This results in a “preferred orientation”. Such “preferred orientation” introduces errors during particle alignment, leading to artifacts in density maps and thereby significantly compromising the quality and interpretability of cryo-EM maps. 

Note: Cryo-EM reconstructs the three-dimensional density map of biological macromolecules using two-dimensional electron beam projection images

On May 16, 2025 (Beijing Standard Time), Junior Principal Investigator Mingxu Hu and Associate Professor Chenglong Bao published a paper in Nature Communications entitled  "CryoPROS: Correcting misalignment caused by preferred orientation using AI-generated auxiliary particles". 

Link to full text: https://www.nature.com/articles/s41467-025-59797-w

The researchers first investigated the reasons for degradation in reconstructed density maps and found it is primarily attributed to particle misalignment caused by preferred orientation. 

Note: A key reason for density map degradation is particle misalignment caused by preferred orientation

To address particle misalignment, the researchers developed CryoPROS, a computational frame that employs deep learning to synthesize auxiliary particles. By combining these AI-generated auxiliary particles with the experimental datasets, the method mitigates the non-uniformity in orientation distribution and corrects estimation bias, thereby enabling accurate determination of orientation angle and precise particle alignment. 

Note: CryoPROS computational frame

CryoPROS achieves a resolution of 3.49 Å with the untilted HA-trimer dataset (EMPIAR-10096), a widely recognized benchmark dataset for preferred orientation. Traditionally, it has been assumed that high-resolution reconstruction would require tilting the specimen stage (an inefficient process that substantially prolongs data collection to acquire the same number of particles), followed by extensive three-dimensional classification, per-particle CTF refinement, and particle polishing. With CryoPROS, the same resolution can now be achieved without these laborious steps. 

Note: CryoPROS achieves a resolution of 3.49 Å with the untilted HA-trimer dataset (EMPIAR-10096), a widely recognized benchmark dataset for preferred orientation

CryoPROS is now available on Github 

This study was led by co-corresponding authors Mingxu Hu (Junior Principal Investigator, Shenzhen Medical Academy of Research and Translation; Researcher, Beijing Frontier Research Center for Biological Structure, Tsinghua University) and Chenglong Bao (Associate Professor, Yau Mathematical Sciences Center, Tsinghua University; Researcher, Beijing Institute of Mathematical Sciences and Applications; Researcher, State Key Laboratory of Membrane Biology, Tsinghua University). Hui Zhang (PhD Candidate, Qiuzhen College, Tsinghua University) and Dihan Zheng (Former PhD Candidate, Yau Mathematical Sciences Center, Tsinghua University) are the co-first authors. This work was funded by the Junior Principal Investigator Starting Fund from Shenzhen Medical Academy of Research and Translation, Beijing Frontier Research Center for Biological Structure (Tsinghua University), and National Key Research and Development Program of China. 

By rapidly freezing samples to trap macromolecules in a life-like state, cryogenic electron microscopy (cryo-EM) generates realistic, high-fidelity images at near-atomic resolution. Mingxu Hu 's team specializes in advancing Cryo-EM technology and methods, and their work has appeared in top-tier academic journals such as Nature Methods, Nature Communications, Communications Biology, Journal of Structural Biology, Nature, Nature Microbiology, and PNAS. 

Mingxu Hu's team is actively recruiting associate researchers, assistant researchers, and postdoctoral fellows in the fields of structural biology, biochemistry, and cryogenic electron tomography (cryo-ET). More details regarding these positions will be announced shortly.