PNAS, 16 November, 2026, DOI:https://doi.org/10.1073/pnas.2620173123
isSPA-vesicle: AI-assisted structure determination of membrane proteins on vesicles
Mingjie Zhao, Duanfang Cao, Ziqi Cheng, Yuhan Hu, Yang He, Shaobo Wang, Xuepeng Wei, and Xinzheng Zhang
Abstract
The demand for three-dimensional reconstruction of membrane proteins in native lipid environments has grown substantially in recent years. Both single-particle analysis and tomographic techniques have enabled structural studies across diverse vesicular samples; however, these methods remain limited in either efficiency or resolution. Here, we present an AI-enabled, vesicle geometry-guided 3D reconstruction strategy based on untilted data. Given that some vesicles exhibit anisotropic shapes in solution and adopt near-spherical or ellipsoidal geometries, we trained an AI model to estimate membrane geometric parameters and compute normal vectors based on the fitted ellipsoidal surface. This information facilitates post hoc filtering of particles picked using the isSPA framework based on the angular difference between their orientations and the membrane normals, and also enables the exclusion of particles outside the vesicles. This allows efficient localization of target membrane proteins enabling near-atomic-resolution reconstruction. Using this approach, we efficiently reconstructed membrane proteins ranging from 130 kDa to 2 MDa across two classes of enveloped viruses and three types of synthetic liposomes, achieving 2.7 Å resolution for influenza A virus hemagglutinin and 3.1 Å for the prefusion spike protein of HCoV-OC43. Our method provides a simple and efficient tool for achieving such reconstructions in physiological lipid bilayers.
文章链接:
https://www.pnas.org/doi/10.1073/pnas.2620173123
相关报道:
附件下载: