Source-linked AI summary
A complete data processing workflow for CryoET and subtomogram averaging
Muyuan Chen, James M. Bell, Xiaodong Shi, Stella Y. Sun, Zhao Wang, Steven J. Ludtke
TL;DR
CryoET provides rich nanometer-scale cellular structural information, but processing tomography data remains labor-intensive and dependent on personnel expertise. The paper presents an integrated workflow that automates processing from tilt-series alignment through subtomogram averaging, achieving high-resolution structures for purified macromolecules and cellular complexes while reducing human effort and increasing throughput.
Problem
CryoET data processing remains labor-intensive and requires substantial computational resources, limiting the efficient study of high-resolution protein structures in cells.
Method
An integrated workflow performs CryoET processing from automated tilt-series alignment and per-particle CTF correction through subtilt refinement and subtomogram averaging.
Results
The workflow achieves state-of-the-art resolutions for purified macromolecules and cells, including 8.5 Å for purified ribosomes and 14 Å for an in situ TolC-AcrAB complex.
Takeaways & Limitations
With a few thousand subtomogram particles, averaged maps at 10-15 Å resolution can be achieved in days, supporting structural annotation and hypotheses about protein identity and composition.
Takeaways & Limitations
The workflow assumes that landmarks are, on average, coplanar with the ice when rotating the X-Y plane for visualization and annotation.
Abstract
from arXiv · showhide
Electron cryotomography (CryoET) is currently the only method capable of visualizing cells in 3D at nanometer resolutions. While modern instruments produce massive amounts of tomography data containing extremely rich structural information, the data processing is very labor intensive and results are often limited by the skills of the personnel rather than the data. We present an integrated workflow that covers the entire tomography data processing pipeline, from automated tilt series alignment to subnanometer resolution subtomogram averaging. This workflow greatly reduces human effort and increases throughput, and is capable of determining protein structures at state-of-the-art resolutions for both purified macromolecules and cells.
Introduction
CryoET enables 3D visualization of molecules in cells and conformationally variable purified molecules, but processing remains labor-intensive despite increased data-collection throughput. The paper presents a complete EMAN2 workflow spanning tilt-series alignment through subtomogram averaging to expedite cellular tomography processing.
- CryoET provides 3D information for isolating overlapping molecular densities in crowded cellular environments.The specimen is tilted within the microscope to provide 3D information about each molecule.
- Despite advances that increase CryoET data-collection throughput, processing still requires substantial human effort and computational resources.
- Cellular tomography data processing has become a major bottleneck in studying high-resolution protein structures.
- The proposed EMAN2 workflow covers the standard CryoET pipeline from raw tilt-series alignment through subtomogram averaging.
Results
The workflow automates CryoET processing from tilt-series alignment and reconstruction through particle identification, CTF correction, de novo modeling, and per-particle-per-tilt refinement. These integrated procedures improve throughput and resolution for purified and cellular samples.
- Automated tilt-series alignment: Tilt-series alignment uses iterative landmark refinement, transform optimization, progressive sampling, and outlier elimination without human intervention.The workflow supports tomograms with or without fiducials.
- Tomogram reconstruction: Direct Fourier inversion with overlapping weighted tiles reduces edge effects and memory requirements while reconstructing tomograms efficiently.The tiles are individually reconstructed and combined using a Gaussian-falloff weighted average.
- CTF correction: Per-particle CTF correction uses full-tilt-image information to estimate defocus robustly, addressing Z-dependent defocus errors in thick specimens.The method maximizes a score over tile-specific defocus values and uses additional constraints at high tilt.
- Initial modeling: De novo SGD modeling generates good initial models within 1 hour on a typical workstation, avoiding dependence on candidate-structure catalogs.The approach begins from an effectively randomized lowpass-filtered map and iteratively updates it using randomly selected particles.
- Subtilt refinement: Subtilt refinement improves purified-ribosome resolution from 13 Å to 8.5 Å and cellular TolC-AcrAB resolution from 19 Å to 14 Å.The procedure independently refines all five orientation parameters for each particle and tilt, compensating for local deviations from global rigidity.
- Scientific application: The integrated workflow enables data-driven cellular structural biology, including structural identification hypotheses and mapping particle positions and orientations back into tomograms.The authors report 10-15 Å averaged maps within days using a few thousand subtomogram particles.
Data availability
The workflow's deposited subtomogram averages include purified ribosome and cellular AcrAB-TolC structures, while the cited figures document alignment, extraction, and refinement stages.
- EMD-0529 is the averaged structure of a purified ribosome, and EMD-0530 is the averaged AcrAB-TolC structure from a cellular tomogram.
- The authors declare no competing financial interests.
- Figure 1 presents the main workflow, tomogram reconstruction, and subtomogram and subtilt refinement workflows.
- Figure 2 documents iterative tomogram alignment and reconstruction for cellular E. coli and fiducialless purified apoferritin data.
- Figure 3 shows particle selection and initial-model generation for TolC-AcrAB pumps, microtubule doublets, and ribosomes.
- Figure 4 compares ribosome averaging before and after subtilt refinement and maps AcrAB-TolC particle locations and orientations back to a tomogram.
Methods
The methods combine coarse-to-fine tilt-series alignment, tiled Fourier reconstruction, stochastic initial-model generation, and per-particle refinement using extracted subtilt series.
- Tomogram reconstruction: Tilt-series alignment begins with coarse alignment and proceeds through iterative landmark and tilt-parameter refinement before final tomogram reconstruction.Landmark coordinates and alignment parameters are refined in multiple rounds.
- Tomogram reconstruction: Landmarks are selected from downsampled, highpass-filtered tomograms using low voxel values, minimum-distance separation, and a default of 20 landmarks.
- Tomogram reconstruction: Overlapping tiled reconstruction uses padding, clipping, and weighted masks to reduce edge artifacts while producing the final tomogram.The tile mask is designed so summed voxel weights equal 1.
- Initial model generation: Stochastic gradient descent generates initial models by repeatedly aligning random particle batches, averaging them, and updating a low-pass-filtered reference.The protocol uses small batches and a learning rate of 0.1 to introduce fluctuations.
- Subtilt refinement: Subtilt refinement searches orientations for each particle and tilt, assesses tilt quality, excludes the worst images, and computes weighted 3D averages.
- Example data processing: Example datasets were processed with automated reconstruction, particle selection, CTF correction, subtomogram refinement, subtilt refinement, sharpening, masking, and FSC filtering.
Supplementary Figure
The supplementary materials describe tiled tomogram reconstruction, per-particle subtilt CTF determination, and program runtime parallelism.
- Fig S1. Tiling strategy: Figure S1 shows padded and clipped individual tiles, overlapping tiles for reduced edge effects, and the resulting by-tile reconstruction.
- Fig S2. Subtilt CTF determination: Figure S2 determines per-particle defoci from tiled tilt-image power spectra, tilt geometry, and particle positions, then uses fitted CTFs for phase flipping.
- Table S1. Program runtimes: Table S1 reports program runtimes and identifies MPI and thread parallelism, including batch-based threading for e2spt_sgd.py.