I could get 3 different nice classes after heterogenous refinement. Below is part of the density:
However, after I selected one of the good classes into homogenous refinement or NU-refinement, I only could get elongated density as below:
I think this is probably something related to the local minimum problem (from 2D classification, it looks all good), could you please give me some suggestions about how to overcome this issue? or by changing the parameter of NU-refinement?
Thank you very much in advance!
What do the 2D classes look like? Because that looks like the effect of a strong preferred orientation…
" …looks like the effect of a strong preferred orientation…"
Alternatively @QChong please show the orientation distribution from homogeneous refinement to confirm.
here are orientation distributions from homogeneous refinement and also NU-refinement.
Almost all of your particles are from a single 2D class, it looks like? This is mostly likely a preferred orientation in the ice problem, but you should examine your 2D classification and make sure you aren’t leaving out good particles from other orientations.
Failing that, you’ll need to screen additives to try and solve the orientation problem.
Hi, thank you all for replying! : )
The interesting thing is that I don’t have orientation-preference issues for another two classes from initial heterogeneous refinement.
- class0, contains 4 subunits, A+B+C+D (based on our current knowledge, alphafold predicted structure)
- class1, contains 3 subunits, A+B+C
- class2, contains 3 subunits, A+B+C
subunit D makes the whole structure become longer (I could see this longer complex side view clearly from 2D). Below are orientations from the initial heterogeneous refinement. I think there is no orientation issue for class 1 and 2, and the strucutre looks nice, which means probably there is no ice issue, otherwise, I guess I should have orientation issues for all complexes.
However, I am still confused why I only have an orientation issue for class0, which just has one more subunit (around 80kDa) compared to others.
You definitely have preferred orientation, per those viewing distribution charts, for all three. It’s just more pronounced for Class 0. Do you have fewer particles for Class 0? It looks like like you do, by eye. That could be why it’s more pronounced.
For reference, here’s what the viewing distribution looks like for a particle I’m currently working on (C1 reconstruction):
Thank you for this suggestion! and I will try to screen grids with additives.
Not necessarily surprising. Very crude way of thinking about it: since your particle is “longer”, it has a harder time being upright in thin-ish ice, so you get less top-views.
I see that’s probably only got side-view from this type of particle!