Separating particles by 3DVA clustering results?

Hi all,

I’m working on a protein which contains both discrete and continuous heterogeneity which can be captured by 3DVA, as seen here:

There are two obvious clusters of particles here, and I’d like to be able to sort/separate the particles by which cluster they belong to, and then use the sorted particles in each cluster to perform a consensus refinement (this would be somewhat analogous to the particle filtering function in cryoDRGN). I was wondering if there was an easy way to do this on cryoSPARC? Also, just in general, is it possible to separate the particles directly by the volume they contributed to generating along a trajectory in the 3DVA job (such that a further consensus reconstruction could be performed on the particles that went into each intermediate volume)? Or, is this further refinement not worthwhile because 3DVA has already done the best alignments/refinements it can against each individual intermediate volume along the way, and redoing refinement on a subset of the particles would not yield any higher resolution map? Lastly, is there a way to visualize where along the plot of the components each individual map came from? Any advice on whether the above are achievable (and if not, if there is a way to mimic the aforementioned functionality in cryoSPARC) would be greatly appreciated.

Many thanks in advance for everyone’s input!

Best,

Adele

Hi Adele,

First off the cryosparc guide has the answers to some of these questions, so that might be worth a look, e.g. the tutorial on “advanced” 3D-VA usage I think should help: Tutorial: 3D Variability Analysis (Part Two) - CryoSPARC Guide

…I’d like to be able to sort/separate the particles by which cluster they belong to, and then use the sorted particles in each cluster to perform a consensus refinement…

You can do this using 3D-VA display, in “cluster” mode. In your case probably you will want to select two clusters. Then you can take the resulting particles in each cluster and refine them separately.

Also, just in general, is it possible to separate the particles directly by the volume they contributed to generating along a trajectory in the 3DVA job (such that a further consensus reconstruction could be performed on the particles that went into each intermediate volume)

Yes, there is an option in 3D-VA display in “intermediates” mode to output particle subsets. Alternatively, you can also perform a 1D clustering operation (also in 3D-VA display), by selecting only a single mode to be used.

Or, is this further refinement not worthwhile because 3DVA has already done the best alignments/refinements it can against each individual intermediate volume along the way, and redoing refinement on a subset of the particles would not yield any higher resolution map?

To my knowledge, 3D-VA is not performing any refinements - it is using the input particle orientations, they are not changing during 3D-VA. Certainly, subsequent refinements of selected subsets after 3D-VA have proved useful.

Lastly, is there a way to visualize where along the plot of the components each individual map came from?

Yes - at least using the “cluster” mode, an interactive 3D-plot will be generated, with each cluster colored differently and numbered as per the outputs. For intermediates mode, a histogram of the mode will be generated, with the component values corresponding to each subset indicated (and volumes will be saved as frame_001, frame_002 etc in the zipped output folder).

Hope that helps!

Cheers
Oli

2 Likes