Use of clustered particles and associated volume from 3DVA

Hi All,

As my sample is quite heterogeneous, I tried 3DVA for resolving confomation heterogeneity.
My question is about the usage of ‘Clustered Particles and associated Volume’, generated by 3DVA->3DVA-display clustered mode. The 3DVA was performed with 270K particles and 3-variability components; and displayed by 3DVA-display cluster mode with 20 clusteres that gave 20 particles-clustes with associated volumes. Could I use the particle-cluster and associated volume directly for homogenous refinement? I already tried this and it gave refined-maps (3.5-3.8 A with approx. 15K particles) showing resonable features of side-chains. Is there any chance of introducing bias with this approach?

Thanks in advance,

Yes, this is an acceptable way to get your final refinement and discrete particle set, as long as you gave a decent lowpass filter for the refinement (or not mess with the default value). You may also try clustering fewer components (eg try component 0 and component 0, rather than all 3 at once) by removing and replacing items in the lower level input.

Hi user123,

thank you for your reply and suggeting of using individual component for clustering. I tried to use individual component by removing particles and volume of other components from lower level input. The job failed with following error: IndexError: index 1 is out of bounds for axis 1 with size 1.

Yes - you have to replace all the existing components with the desired component. So if you have four components in the initial 3D-VA job, you will need to provide four copies of e.g. component 1, for example.

1 Like

Thank you for the suggestion.

I ran 3DVA with 3 components, then ran 3DVA analysis cluster mode with just 2 copies of the same component. Is this ok? I never tried with 3 copies of the same component.

should be fine if it works! In my hands if I don’t have the same number of components as the input cryosparc throws an error. @team it would be great if clustering using a subset of components was “officially” supported, I have found it very useful for isolating low occupancy states