/ cryosparc tips

Heterogeneous Refinement in cryoSPARC


Heterogeneous Refinement simultaneously classifies particles and refines structures from n initial structures, usually obtained following an Ab-Initio Reconstruction. This facilitates the ability to look for small differences between structures which may not be obvious at low resolutions, and also to re-classify particles to aid in sorting.

When to use Heterogeneous Refinement

You suspect heterogeneity in your dataset (i.e., a dataset containing more than one structure) on which you have already performed a multi-class Ab-Initio Reconstruction in cryoSPARC and obtained n different coarse-resolution structures, or obtained n initial models from elsewhere. You now want to examine (possibly a subset of) the data to see if there are additional structures present that are similar but have smaller differences from the structures you have already obtained.

Looking for heterogeneous classes/structures starting from initial models

Select the dataset of interest from the Datasets page, then navigate to the Experiments page. Locate the result of a previous Ab-Initio Reconstruction in cryoSPARC, the result of a previous refinement, and/or an Imported Volume which you previously uploaded. (For instructions on how to import an initial model, please refer to this post.)

In this example, we had previously performed a 5-class Ab-Initio Reconstruction in cryoSPARC on the dataset of interest. We now wish to see if the 5th class, which contains 38.2% of the total number of particles (as indicated by the orange flag on the top left of the projections), can be further separated into more classes which are slightly different from each other. Therefore, in this Heterogeneous Refinement experiment, we selected the Particles corresponding to the 5th class, and we decided to try and separate those particles into three different classes, by selecting the Structure associated with the 5th class, three times. (It is not necessary to start with three classes; this is simply an example.)

Alternatively, we could have selected three different structures as the initialization for this Heterogeneous Refinement, as in the example below.

It is also possible to select Structures from the result of different experiments, or from an Imported Volume.

Once you have selected the Particles of interest and the number of Structures you wish to find, choose New Experiment > Heterogeneous Refinement. Should you wish to change any of the default parameters, do so from the Setup page.

Click Launch and Enqueue to commence the Heterogeneous Refinement. Once complete, you can view the result on the Experiments page.

On examining the result, you may wish to:

  • Perform one or more additional Heterogeneous Refinement experiments, using different starting models or a larger number of starting models, to further separate particles into classes that are slightly different from each other

  • Select a particular structure and set of particles and perform a Homogeneous Refinement in order to obtain a higher resolution refined structure along with gold-standard FSC resolution estimates.

  • Mix-and-match processing steps including 2D classification, Ab-initio reconstruction, and heterogeneous refinement to optimally sort and partition your particle sets.

Check out these related posts on sub-classification and rapid 2D classification in cryoSPARC.

Questions or feedback? Visit the cryoSPARC discussion forum or send us an email at feedback@structura.bio.

Using cryoSPARC for the first time? Find out how to install and set up cryoSPARC for best results.