GRECO: Geometric Registration of Electron Cryotomography Observations

Tomographic reconstruction is challenging because of limited angular coverage, noise, specimen motion and radiation damage. Here, I present GRECO, Geometric Registration of Electron Cryotomography Observations, an algorithm that uses local specimen structure to fit a shared three-dimensional geometry. Starting from central views, it selects landmarks from a working reconstruction and progressively adds neighbouring views. Near zero tilt, uncertainty in landmark depth has less effect on the predicted image position, allowing provisional landmarks to guide the first steps. Each landmark contributes through views with useful local information. For each observation, GRECO reconstructs a local reference from the other supporting views and compensates for the directional blur produced by those viewing angles. A joint fit updates image translations and landmark positions, linking overlapping angular tracks across the series. GRECO produced reconstructions suitable for subtomogram averaging, detection of small macromolecular complexes and detailed membrane segmentation across specimens ranging from vesicles with fiducials and a fiducial-free mixture to an intact near-minimal cell and cellular lamellae imaged in situ. The Greco software combines this registration algorithm with fast preprocessing, reconstruction and TinyCARE denoising, providing a practical workflow for studying molecular structure and cellular organization.