摘要

In single-particle electron microscopy (EM), multiple micrographs of identical macromolecular structures or complexes are taken from various viewing angles to obtain a 3D reconstruction. A high-quality EM reconstruction typically requires several thousand to several million images. Therefore, an automated pipeline for performing computations on many images becomes indispensable. In this paper, we propose a modified cross-correlation method to align a large number of images from the same class in single-particle electron microscopy of highly nonspherical structures, and show how this method fits into a larger automated pipeline for the discovery of 3D structures. Our modification uses a probability density in full planar position and orientation, akin to the pose densities used in Simultaneous Localization and Mapping (SLAM) and Assembly Automation. Using this alignment and a subsequent averaging process, high signal-to-noise ratio (SNR) images representing each class of viewing angles are obtained for reconstruction algorithms. In the proposed method, first we coarsely align projection images, and then realign the resulting images using the cross correlation (CC) method. The coarse alignment is obtained by matching the centers of mass and the principal axes of the images. The distribution of misalignment in this coarse alignment is estimated using the statistical properties of the additive background noise. As a consequence, the search space for realignment in the CC method is reduced. Additionally, in order to overcome the false peak problems in the CC, we use artificially blurred images for the early stage of the iteration and segment the intermediate result from every iteration step. The proposed approach is demonstrated on synthetic noisy images of GroEL/ES. Note to Practitioners-This paper concerns the automated alignment of the large number of noisy images that must be handled when class averaging is applied in single-particle electron microscopy. The new proposed method consists of prealignment, iterative alignment using the CC, artificial image blurring and image segmentation. The prealignment is obtained by matching the center of mass and the principal axis of the images. This results in a SLAM-like distribution of pose with quantifiable covariance, on which computations can be performed. Next the prealigned images are aligned more accurately through the iterative CC method with image blurring and segmentation. The most notable improvement is the prealignment step. Although this prealignment inherently results in imperfect alignment because of the background noise on the images, the statistical information of the imperfect alignment can be obtained and is used for the iterative CC at the next step to obtain better alignment at the end. Since the prealignment involves the principal axes of images, the alignment method proposed in this paper targets the alignment of non-circular projection images.

  • 出版日期2014-7