Ab-initio reconstruction in cryoSPARC
cryoSPARC allows for rapid, unbiased single- and multi-class reconstructions from both homogeneous and heterogeneous datasets, without any initial references or prior structural knowledge.
Figure 1 provides an example of a typical ab-initio experiment workflow starting from a new dataset. (To enlarge the image, you can open the image in a new tab or download it.)
A. When you upload a new dataset into cryoSPARC, it is recommended to perform a single-class (i.e., homogeneous) ab-initio reconstruction as the first step, in order to better understand the data. There is no need to input an initial model into cryoSPARC for this experiment.
To do so, select your dataset from the Datasets page, then navigate to the Experiments page. On the right hand side below the 'New Experiment' button, you will notice a 'Selection' box which contains a field to input the number of ab-initio classes, which by default is set to 1. For this experiment, there is no need to change the default value. Click on 'New Experiment' and select 'Ab-initio'. You can adjust any other parameters if desired, but there is no requirement to do so. Click 'Launch' and 'Enqueue' to commence the experiment.
The result of the experiment can be viewed from the Experiments Page as an image projection. If you are satisfied that there is only one class present in the data, then you can proceed to refinement (to be covered in cryoSPARC Workflow Series: Part II).
B. If you wish to further explore the data, it is recommended to perform an additional ab-initio reconstructions, this time with number of ab-initio classes > 1 (i.e., heterogeneous ab-initio reconstruction).
To do so, start from the 'Experiments' page. This time, in the 'Selection' box, enter the number of ab-initio classes you wish cryoSPARC to use (e.g., 5). Then click 'New Experiment' and 'Ab-initio', followed by 'Launch' and 'Enqueue' to commence the experiment.
The resulting classes can be viewed from the Experiments page, while the experiment is in progress. On completion, you can view the proportion of particles that fell into each of the classes you asked cryoSPARC to use.
C. It is recommended to perform further iterations of heterogeneous ab-initio reconstruction in order to explore the results with different numbers of classes. To do so, perform the same steps as in B., but ensure to change the number of ab-initio classes in the 'Selection' box each time. We typically suggest trying between 3 and 8 classes on every dataset to enable discovery of different structures.
The results of ab-initio reconstruction can vary widely depending on characteristics of the dataset. It can be important to examine pose distribution plots of results (which can be viewed and downloaded from the live result stream on the Launch page), as well as class distributions.