Support for large high symmetry (e.g. viruses) particles in Ab-initio and Hetero-refinement

Features and Functionality
1

Hi,

I have noticed with more than 10 different datasets of virus samples (I symmetry) that both
Ab-initio reconstruction and Heterogeneous Refinement never result in proper 3D densities
when I symmetry is not enforced. Instead, we always get donut/ring-shaped volumes. Enforcing
I symmetry however does result in correct volumes. I am afraid that enforcing symmetry at this
early stage of processing leads to sub-optimal final results. Also, we do try all the tips & tricks
recommended on this board and elsewhere for dealing with symmetries (e.g. increasing batch
size or initial/final resolution etc.), but so far this has not solved our issues in any of our datasets.

I am wondering if there are any plans to introduce new features or optimize workflows for
dealing with high symmetry particles?

Thanks!

2

I wonder if it’s mostly a sampling issue. If you process these data in Relion, is 7.5 deg. initial sampling sufficient? Also, does homogeneous refinement with some known good particles also fail in C1? If you use the I symmetry classification to throw away junk/non-particles, do these methods continue to fail in C1?

3

@twg thanks for reporting this.
In addition to the questions from @DanielAsarnow do you notice that if you do ab-initio reconstruction with I-symmetry enforced, and then use that initial model for heterogeneous refinement, do you still end up with donut/ring shaped volumes?

4

Sorry for the delayed reply, here are my answers @DanielAsarnow @apunjani:

If you process these data in Relion, is 7.5 deg. initial sampling sufficient?
-Relion always gives me proper closed shells and not donuts with 7.5 degrees sampling

Also, does homogeneous refinement with some known good particles also fail in C1?
-Yes, homogeneous and heterogeneous refine have never given me a closed shell even using
particles that are definitely good

If you use the I symmetry classification to throw away junk/non-particles, do these methods continue to fail in C1?
-Yes, as stated above, refinements without I enforced never give me closed shells, independent of the initial volume or particles

Do you notice that if you do ab-initio reconstruction with I-symmetry enforced, and then use that initial model for heterogeneous refinement, do you still end up with donut/ring shaped volumes?
-Even using excellent initial models (either from ab-initio with I enforced or even from published I symmetry maps), when running C1, I will always get donuts

Not really sure what the issue is. As mentioned above, this is the case for all icosahedral samples I have tried to process in C1 with CryoSPARC.

5

Here are some example outputs:

These are 2D class averages:
image

And here is an example output from hetero refine in C1:
image

The initial model used was of course a closed shell icosahedral density from a related capsid.

Here is what I have already tested:
-lowering Initial Resolution
-increasing Batch Size per class

Both don’t seem to make a difference.

6

Hi cryoSPARC team,

It seems this issue still persists. Is there any plan to address these problems at any point?
It is very unfortunate that both ab-initio and hetero refine simply cannot handle any I symmetry
particles we try in C1. We have literally tried dozens of different I symmetry particles (mostly
viruses) and none of them work (all of them become donuts). And as stated above, using Relion
always works. We would love to be able to just stick with cryoSPARC instead of having to switch
back and forth. Any update on this issue would be appreciated.

Best,
twg

7

Just giving this a bump. Any updates from the cryoSPARC team? @apunjani

Best,
twg

8

I think this is a similar failure mode to what we see here:

And I wonder if it could be addressed with some kind of prior (e.g. seeding with spherical density).

I vaguely recall that in cryosparc v0 there used to be an option to do exactly this - @team can you confirm or is my memory failing…?

1 Like