Alignment of cryo-EM movies of individual particles by optimization of image translations

Abstract

Direct detector device (DDD) cameras have revolutionized single particle electron cryomicroscopy (cryo-EM). In addition to an improved camera detective quantum efficiency, acquisition of DDD movies allows for correction of movement of the specimen, due to both instabilities in the microscope specimen stage and electron beam-induced movement. Unlike specimen stage drift, beam-induced movement is not always homogeneous within an image. Local correlation in the trajectories of nearby particles suggests that beam-induced motion is due to deformation of the ice layer. Algorithms have already been described that can correct movement for large regions of frames and for >1 MDa protein particles. Another algorithm allows individual <1 MDa protein particle trajectories to be estimated, but requires rolling averages to be calculated from frames and fits linear trajectories for particles. Here we describe an algorithm that allows for individual <1 MDa particle images to be aligned without frame averaging or linear trajectories. The algorithm maximizes the overall correlation of the shifted frames with the sum of the shifted frames. The optimum in this single objective function is found efficiently by making use of analytically calculated derivatives of the function. To smooth estimates of particle trajectories, rapid changes in particle positions between frames are penalized in the objective function and weighted averaging of nearby trajectories ensures local correlation in trajectories. This individual particle motion correction, in combination with weighting of Fourier components to account for increasing radiation damage in later frames, can be used to improve 3-D maps from single particle cryo-EM.

6S RNA Mimics B-Form DNA to Regulate Escherichia coli RNA Polymerase

“…Poorly populated classes were removed, resulting in a dataset of 808,484 particles. These particles were individually aligned across movie frames using direct-detector-alignjmbfgs software (Rubinstein and Brubaker, 2015). Aligned particles were 3D auto-refined in RELION using a model of Eco Eσ 70 (PDB ID 4LK1) (Bae et al, 2013) low-pass filtered to 60 A resolution using EMAN2 (Tang et al, 2007) as an initial 3D template.…”

Overview and future of single particle electron cryomicroscopy

“…The motion was found to be greater during the early part of the low-dose exposure, decreasing as the specimen stabilised. Most recently this has led to a number of improved protocols for image processing that effectively deblur the movie frames [19,40,41,4,28,32]. Fig.…”

“…The first type of beam-induced motion can, in principle and to a great extent in practice, be corrected by subsequent computerbased image processing to align the individual frames of the movies collected using the new CMOS detectors [19,40,41,4,28,32]. The second type of pseudo-Brownian motion cannot be avoided but fortunately seems to be small enough [25] to be of minor significance for the practice of single particle cryoEM structure determination.…”

High resolution single particle refinement in EMAN2.1

“…By aligning the frames before averaging to produce a ‘micrograph’, it is often possible to eliminate the majority of the drift which was previously a large problem[12]. There are additional variants of this methodology which focus on aligning the individual particles within a micrograph, since they may not all move in a uniform way [13,14]. There are a number of software packages for performing whole-frame alignment [15,16], and this issue remains an active area of development.…”