Great 3D variablity output, poor homogeneous refinement output

When I use 3D variability to do analysis of movements of my protein, I get an output of a map that represents “component_0” that is amazing. However, when I attempt to refine my protein with other types of refinement such as homogeneous refine or non uniform refine I get very messy outputs of my structure that have a lot of features that don’t make sense.

Originally, I thought it could be that the output was from a particular subset of particles but the 3D variability job is using the full particle set inputted to produce this map. This map isn’t an intermediate it is the map produced from the actual 3D variability job not the display job. I reimported the particle set back into cryosparc with the alignment parameters from the result output particle file from the 3D variability, but it did not improve the results of refinement.

My question is what could be causing the great output from the 3D variability job that is not being done by the homogeneous refinement or non uniform refinement. I’ve used both of the other refinements before successfully so I’m a bit confused as to why this could be happening.

@apunjani or anyone that may have a similar experience do you have any advice as to how I can replicate the 3D variability output in other refinements, or do you have a potential explanation as to why this might happen?

Hi @adtaheri,

To confirm, do you mean that you opened the cryosparc_PXX_JXX_component_0.mrc
file from a 3D Var job, rather than the volume series from a 3D Var Display job?

This file is actually the variability component, rather than an actual density map from a certain set of particles - it represents the change in density vs. the consensus refinement that was used as input for 3DVA.

Can you try using the 3D Var Display job, set the mode to ‘intermediates’ ? It will perform reconstructions from subsets of the particles. See if any of the reconstructions along component 0 look like a better resolved version of your protein.

Thank you for the response.

When I use the 3D Var Display job with intermediates I get approximately half of the frames being very good reconstructions of the protein. I realize that the particle subsets from these reconstructions can be non overlapping by setting width to a certain parameter. However, the .csg file outputted refers to a series as a whole and I’m not quite sure how I can separate the intermediates. Any advice at all there would be great. I saw in another thread importing the result group can work but how would you recommend moving from there.