Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2016 May 1;100:25-34.
doi: 10.1016/j.ymeth.2016.02.018. Epub 2016 Feb 27.

High Resolution Single Particle Refinement in EMAN2.1

Affiliations
Free PMC article

High Resolution Single Particle Refinement in EMAN2.1

James M Bell et al. Methods. .
Free PMC article

Abstract

EMAN2.1 is a complete image processing suite for quantitative analysis of grayscale images, with a primary focus on transmission electron microscopy, with complete workflows for performing high resolution single particle reconstruction, 2-D and 3-D heterogeneity analysis, random conical tilt reconstruction and subtomogram averaging, among other tasks. In this manuscript we provide the first detailed description of the high resolution single particle analysis pipeline and the philosophy behind its approach to the reconstruction problem. High resolution refinement is a fully automated process, and involves an advanced set of heuristics to select optimal algorithms for each specific refinement task. A gold standard FSC is produced automatically as part of refinement, providing a robust resolution estimate for the final map, and this is used to optimally filter the final CTF phase and amplitude corrected structure. Additional methods are in-place to reduce model bias during refinement, and to permit cross-validation using other computational methods.

Keywords: 3-D reconstruction; CryoEM; Image processing; Single particle analysis; Structural biology.

Figures

Figure 1
Figure 1
Per particle SSNR estimation. A. Exterior mask. B. Central mask. C. Raw unmasked particles. D. Background region. E. Particle region. F. Power spectra for each masked particle stack compared to traditional background estimate (dotted line), which underestimates noise and overestimates signal. SSNR is computed from the two masked curves. In this Beta-galactosidase example, there is only a modest difference between the two background calculations. In specimens with detergent or continuous carbon, the impact on SSNR estimation can be as much as an order of magnitude[33].
Figure 2
Figure 2
An overview of the iterative processing strategy implemented for reference-free class-averaging in e2refine2d.
Figure 3
Figure 3
Reference-free class-averages used to produce an initial model using Monte-Carlo method implemented in e2initialmodel. Two of the possible 3-D starting maps are shown on the left along with corresponding class-average projection pairs for comparison on the right. For a good starting model, projections (1st and 3rd columns) and class-averages (2nd and 4th columns) should agree very well. The lower map exhibits poor agreement, so the higher ranked upper map would be used for 3-D refinement.
Figure 4
Figure 4
Overview of automatic refinement process described in the text and implemented in e2refine_easy. The process begins by splitting a user specified set of particles into even (white) and odd (grey) sets. Following the “gold standard” protocol, the initial model is phase randomized twice at resolutions higher than ~1.5x the target resolution. The two perturbed starting maps are then refined independently against the even and odd halves of the data. Iterative refinement begins by reprojecting the initial map and classifying particles according to their similarity to these projections. Classified particles are then iteratively aligned and averaged as shown in Fig. 2 (lower). The resulting class averages are then reconstructed in Fourier space to form a new 3D map, which becomes the starting map for the next iterative cycle. At the end of a user-specified number of iterations (typically 3–5), the process terminates. A Fourier shell correlation is computed between all pairs of maps produced from refining the even and odd subsets to assess resolution at each iteration and monitor convergence. In the final step, the even and odd maps are averaged, CTF amplitudes are corrected and the FSC is used to create a Wiener filter, ensuring that only the consistent portions of the separate refinements are visualized in the final averaged map.
Figure 5
Figure 5
Refinement results of the Beta-galactosidase test data subset from the EMAN2.1 tutorial. e2refine_easy was run 4 times sequentially in this test, and the final FSC curves from each run are combined in one plot. The first 2 runs used downsampled data for speed, so the FSC curves do not extend to as high a resolution. The inset shows that beta-strands can be clearly resolved and alpha helices have appropriate shape. Equivalent results could have been achieved in a single run, but the intermediate results are useful in the context of the tutorial, and require less compute time. The table describes the basic parameters and wall-clock time of each refinement run. The final run was performed on a Linux cluster using 96 cores (~250 CPU-hr).

Similar articles

See all similar articles

Cited by 34 articles

See all "Cited by" articles

Publication types

LinkOut - more resources

Feedback