Case Study: Processing EMPIAR-10291 (300 Micrographs) to 3.4Å in 1 hour 25 minutes

February 2020

Structure of an Undocked Hemichannel of the N-terminal-deleted INX-6 in a Nanodisc

Three structures from a recent paper in Science Advances (Burendei et. al, Science Advances 2020) depicting the Cryo-EM structures of undocked innexin-6 hemichannels in phospholipids were recently released on EMPIAR, Electron Microscopy Public Image Archive:

EMPIAR-10289 in light blue, EMPIAR-10290 in green, EMPIAR-10291 in light yellow

This case study focuses on EMPIAR-10291, which contains 300 motion-corrected micrographs. In cryoSPARC, you can resolve a 3.4Å structure (published resolution of 3.6Å) with no manual picks and little configuration in less than 90 minutes:

Workflow in cryoSPARC

Iterative Workflows in cryoSPARC

The primary benefit of cryoSPARC's speed is the ability to quickly iterate through the cryo-EM data processing pipeline and experiment along the way. It is recommended to perform a first-pass workflow from raw data through to a refined structure (as outlined above) to get a sense of the quality of your data before (or concurrently with) proceeding to optimize various stages of the processing pipeline. The quality of a reconstruction is dependent on optimizing various stages of the pipeline:

1) Pre-processing (Exposure Curation, Patch CTF Estimation)

  • Exposure curation can assist with filtering exposures with poor CTF fits or bad ice; this helps increase the quality of particle picks

  • Generally, CTF estimation auto-tunes parameters based on the input data and does not require tweaking. It also handles tilt data directly with no changes.

2) Particle curation (picking, 2D classification)

  • Experimenting with the minimum and maximum particle diameter parameters in the Blob Picker job in combination with different particle extraction box sizes in 2D Classification to better the quality of 2D Classes. Generally, the particle should be half or less the width of the box.

  • We experimented with using refined volumes (after getting a first reconstruction) and the "Create Templates" job to generate 2D projections and feed those into a "Template Picking" job to improve picks. This was found not to yield a higher resolution result because the blob picker already did very well.

3) Reconstruction and refinement variations

  • In this scenario, a single class ab-initio resulted in the best input for the refinement

  • Multi-class ab-initio can help to filter particles in addition to 2D Classification

  • Multi-class heterogenous refinement is also useful for pruning outlier particles in later stages of processing. This can be started from multiple ab-initio volumes.

  • 'New Homogeneous Refinement' is recommended as it features many performance enhancements and the ability to perform on-the-fly CTF refinement versus the Legacy Refinement. We were able to complete the first refinement in under 10 minutes on a single GPU.

  • In this case, we found that CTF refinement did not help in the final reconstruction, as the particle is a membrane protein and so the disorder in the micelle makes it difficult for CTF refinement to correctly estimate the defocus or higher order aberrations present during imaging.

  • Non-uniform refinement was used due to the micelle surrounding the target. This provided a small improvement in structure resolution, but did take 1 hour instead of the 10 minutes for a standard homogeneous refinement.

Below is an example of this iterative workflow in action:

Multiple rounds of the New Refinement and Non-Uniform Refinement were conducted on different ab-initio classes.