Job: Heterogeneous Reconstruction Only

At a Glance

Produce a 3D volume for each class from particle images that were previously classified in an upstream job, using the images’ existing pose estimates and class assignments.

Description

Heterogeneous Reconstruction Only produces a map for each class from a particle stack, using the particles’ existing pose estimates and class assignments. No alignment is performed. This is most often used to produce full-size maps of classes from jobs like Heterogeneous Refinement, which downsample particle images by default.

Jobs which align and classify particles (like Heterogeneous Refinement and Ab-Initio Reconstruction) do not perform gold-standard half set alignment. Thus, although Heterogeneous Reconstruction Only produces FSC curves, they may overestimate the maps’ resolution. Particles from each class can be refined in a job like Homogeneous Refinement which does use half set alignment before peforming any tasks for which resolution estimates are important (for instance, model building).

Inputs

Particles (all classes)

One or more “all particles” particle stacks which will be reconstructed. These particles must have pose and class estimates (alignments3D_multi slot). Note that all particles should have poses and class assignments for a common reference (e.g., they should all come from the same refinement or classification). If particles come from different classification jobs, they will be mixed together and produce unexpected results.

You must use the “all particles” output of the upstream job (e.g., Heterogeneous Refinement, Ab-initio Reconstruction, 3D Classification, etc.), which contains information about the particles’ probability of belonging to each class. If you want to reconstruct a subset of classes, using Homogeneous Reconstruction Only once for each class may be preferred.

Static mask

If provided, this mask will be used for FSC calculation. FSC masks are typically tighter than refinement masks, so the Resolution Mask is typically the correct choice from among the masks produced by an upstream refinement.

If not provided, a mask is generated using the same procedure as in a refinement. This generated static mask can optionally be auto-tightened by turning on Optimize FSC mask.

Commonly Adjusted Parameters

Reconstruction box size (voxels)

Box size of the reconstructed volume. By default, volumes are produced with the same box size as the input particle stack.

Force hard classification

By default, a particle’s contribution to a class is proportional to its probability of belonging to that class. If Force hard classification is turned on, particles only contribute to the map of their most likely class.

Note that if the upstream classification was performed with hard classification turned on, the particles will already have probability 1.0 in the most likely class and probability 0.0 of belonging to any other class, so this setting will have no effect.

Symmetry

This symmetry will be applied to the particles during reconstruction. Note that the particles are not aligned to a symmetry axis, so if the previous poses were not found with enforced symmetry this setting may produce unexpected results. See Symmetry in CryoSPARC for CryoSPARC's symmetry conventions.

Helical symmetry

If the input particles are helical, then Helical twist (degrees), Helical rise (Angstroms), and Helical symmetry order to apply must all be entered. These values can be found in the Helical Refinement job card or at the bottom of a Helical Refinement job’s Event Log.

Force re-do half-set split

If this parameter is turned on, particles will be re-assigned to half sets before the volumes are reconstructed. If particles do not have half-set assignments they will be split regardless of this setting.

Flip the reconstruction hand

During the early stages of a refinement, there is no way to determine the handedness of particle images. Maps may therefore be the mirror image of the true volume.

Turning on Flip the reconstruction hand reassigns particle poses such that they match the mirror image of the original reference. This has the effect of reflecting the maps along the Z-axis and adjusting the particle poses to match.

Ignore higher-order CTF parameters

Ignore tilt, Ignore trefoil, and Ignore tetra each turn off the indicated higher-order CTF correction. This is a quick way to assess the relative contributions of each of these corrections to the final map quality.

Ewald sphere correction

The Ewald sphere (EWS) corrects for the fact that cryo-EM particles have thickness and, thus, slightly varying defocus along the Z-axis. In general, this correction is only necessary for large objects which refine to very high resolution. More information is available in the Ewald Sphere Correction tutorial.

Do EWS correction

If this parameter is turned on, the projection will correct for the EWS.

EWS Curvature sign

The EWS can have either positive or negative curvature. There is no practical way to know which is correct a priori. In general, we recommend that EWS curvature is determined for a single class using Homogeneous Reconstruction Only.

Optimize FSC mask

If this parameter is turned on, the FSC mask’s tightness is optimized to find the best GSFSC resolution, as in Homogeneous Refinement or Non-Uniform Refinement.

Wide mask final output

If this setting is turned on, a wide mask will be applied to the map. This sets many voxels to exactly 0, which makes the compressed size of the file smaller. The uncompressed file (i.e., the .mrc file that you download) will be the same size regardless of this parameter’s value. The magnitude of this effect depends on how much of the box is occupied by solvent.

Because web browsers download large files using compression, turning this parameter on also makes volume downloads faster.

Output zipped half-maps

By default, Heterogeneous Reconstruction Only saves only unsharpened and sharpened maps for download. Turning this parameter on also saves each class’s half maps. Note that some classification jobs (notably Heterogeneous Refinement) do not obey gold-standard half set independence, so care must be taken when using these half maps downstream.

Outputs

Particles

The input particles. If Flip the reconstruction hand is turned on, the particle poses will be adjusted to match the flipped volume. Otherwise, they are unchanged (Force hard classification is only used for reconstruction; the particle probabilities remain the same as the input regardless of this parameter setting).

All volumes

A single Volumes group containing the reconstructed volumes.

Recommended Alternatives

If reconstructions of a subset of a job’s classes are needed, using a Homogeneous Reconstruction Only for each desired class may be more desirable.

We recommend using Homogeneous Reconstruction Only on a single refinement’s particle stack to determine the Ewald sphere curvature.

Note that most jobs in CryoSPARC begin with a downsampled map, so reconstructing at the full box size is usually not necessary except for visual inspection/presentation.