Resolving merged 2D classes

Hi everyone,
I am currently processing data for a soluble protein that looks like a mushroom. (please refer to very-rough sketch below) I have expected to have single-stem 2D classes, however, have been getting dual or sometimes trio-stem. However, the top part of the stem perfectly aligns in the 2D class, only the stem part is fuzzy/ I suspect that I am having multiple states of my protein.
How should I go about to group my particles into single states?

I have tried to extract particles from 3 single classes and further 2D class them individually, but still got multiple stems in 50 classes. The particle count for each class ranges from 9,6k to 17,8k.

I can try 3D classification, but so far due to the preferred orientation problem of my protein, the whole length of the protein cannot be reconstructed in 3D. The multiple states maybe also adds to this problem.

Does anyone have any ideas to resolve this issue? I am looking forward to a fruitful discussion!

Thank you.

1 Like

Is your protein symmetric in such a way that the top view and the bottom view of the protein result in identical projections minus the stem?

It sounds like you have motion of the “stem” relative to the rest of the protein? There’s not really much you can do about this in 2D. How many total particles do you have?

If you have a decently large particle set, get a rough consensus reconstruction using homogeneous or NU refinement (after running multi-class ab-initio), then use 3DVA to try and peel apart your structural conformers. If you’re lucky you may be able to use the clustering mode of 3DVA analysis to get more conformationally uniform particle stacks. You may have to tinker with masking the different regions of your protein to accomplish this.

Is the “stem” truly blurry, or does it seem to have two/three distinct positions? If the former, you’re probably dealing with continuous flexibility which is going to be tough to resolve — maybe 3DFlex or 3DVA could help as @gdodge suggests. If the latter, 3D classification could probably help more than 3DFlex/3DVA.

I have had a similar issue where a significant portion of my particle is conformationally stable (~120 kDa), with a flexible domain (~15 kDa) on the side showing up as a blur in 2D classes and lower resolution in 3D refined maps.

The most success I had (after multiple 2D class rounds, ab initio, het refine, etc.) was masking the stable portions of my proteins and running a few local refinements on that region to get the best alignment. From there, 3DVA masking the flexible domain supported continuous motion to this region (as opposed to discrete classes).

Dear @kyestachowski ,

Yes it is. The direct top and bottom view are very hard to distinguish in 2D, as well as slightly tilted ones. So I have to look very closely to choose the side views to try to resolve the issue

I have around 700,000 particles, but only about 30% of them are good side views. I will try using 3DVA as you sugegsted. Thank you!

Hi @posertinlab ,

Thank you for the suggestion! I have tried 3D classification and it does not resolve the issue, as in all resulting classes cannot reconstruct that bottom part properly. I guess I have both issues here, as in the stem is blurry and there could be at least 3 distinct positions of the protein as well.
I will try my hand at 3DFLex/3DVA. Thanks a lot!

I would wager that you are facing two problems. The stem itself is flexible and top/bottom views are being aligned together. In any case, I think the above advice (3D classification/3DVA/3DFlex) is probably the only way to tackle this.

Sticking with your mushroom terminology:

What is the best resolution you can achieve of the mushroom cap from a NU-refinement?
What is the MW of the stem and the cap?