Search papers, labs, and topics across Lattice.
This study introduces a novel approach to reconstruct protein conformations from single-particle cryo-electron microscopy (cryo-EM) data by directly modeling the protein backbone as an atomic point cloud. The method employs an indirect shape matching framework where a point-cloud template is iteratively deformed using gradient flows on a Lie group until its simulated projections align with the noisy data. The results demonstrate the capability to accurately recover both single- and multichain proteins and effectively capture conformational transitions, showcasing significant advancements in cryo-EM reconstruction techniques.
Directly reconstructing protein backbones from cryo-EM data without intermediate electrostatic maps reveals new potential for capturing complex conformational changes.
Single-particle cryo-electron microscopy images a macromolecule as many noisy tomographic projections of its electrostatic potential. We reconstruct the protein backbone directly from such projections, as an atomic point cloud, without the intermediate step of reconstructing the 3D electrostatic potential map. We formulate this as an indirect shape matching problem: a point-cloud template of the backbone is deformed until its simulated projections agree with the data, with the structure observed only through the imaging operator. The deformation is computed via a gradient flow on a Lie group, and we derive the resulting framework in a general geometric setting before adapting it for single-particle cryo-electron microscopy. On synthetic data, we recover single- and multichain proteins and capture conformational transitions.
