I have an icosahedral viral capsid that is a little too flexible to resolve well with I symmetry, so I am trying to extract the 12 pentasymmetrons and refine them separately, ideally with C5 symmetry applied. This is turning out to be less straightforward than I expected.
My process is as follows:
Extract 960 px full-resolution particles containing the entire capsid
Homogeneous refine with I symmetry applied
Expand symmetry (60 x original particle number)
Remove duplicates with same center (1/5 particle number from 3, 12 x particle number from 1)
Use segger in chimera to create masks of 1 pentasymmetron, and 11 other pentasymmetrons/rest of the capsid. Import these into CS
Use Vol Align Tools to recenter and rotate the 1 pentasymmetron mask from 5, the entire capsid volume map from 2, and the particles from 4 such that the chosen pentasymmetron is centered and has its 5-fold axis aligned to the Z axis.
Use a Vol Align Tools job with the same center and rotation as 6 on the 11 pentasymmetron/rest of the capsid mask from 5 and the entire capsid volume map from 2
Particle Subtraction using the aligned particles from 6 and the volume and mask from 7.
Downsample particles with Recenter using aligned shifts
Local refine using particles from 9, mask from 6
However, neither the Particle Subtraction nor the Downsample jobs appear to be properly centered.
A similar problem was discussed here: How to obtain the portal vertices coordinates of an icosahedral virus
and a similar protocol was found to work. However, they did not attempt to rotate the subparticles, or substract the particle signal, instead re-extracting boxes directly from the micrographs. I tried extracting the particles from step 6 in my protocol and the results of Homogeneous Reconstruct Only show the particles are not centered the way I expected from the Vol Align Tools results:
I had been able to get expected results from Particle Subtraction and Downsample jobs when I used Vol Align Tools with recentering but not rotation, so I believe the rotation is introducing the problem, but reading the guide page for Vol Align Tools, I donât see why that should be.
Here are the Euler angles entered in the ârecenter_axangâ field of the Vol Align Tools job:
0.884226, 0.553461, 1.5710
I am using Cryosparc v4.7.1
Any help or thoughts would be appreciated.
EDIT - I have found that the Symmetry Expansion job also moves the center of the particles by an unknown amount, and is likely the source of the unexpected offset. I will continue to experiment, and update here if I figure everything out.
When you extracted after vol alignment tools, did you select the option to recenter particles using alignments? Otherwise the results will not be centered, as you observe
EDIT: although having said that, in that case I would expect the original centering, not so far off centerâŚ
The sym expansion step should be prior to vol alignment tools - did you perhaps perform it afterwards?
Extraction (and box size cropping in the Downsample job) were done with the ârecenter particles using alignmentsâ option checked. I tried running extraction at several points after sym expansion with that option unchecked, and all boxes were centered on the center of the entire capsid, as expected.
Sym expansion is indeed before vol alignment, the particles input to the Sym Expansion job are the direct output of a homo refine job with I symmetry applied. Here is a homo reconstruct only (C1) job run with the output of the Remove Duplicates job described in step 4 of my protocol:
So the center has been moved away from the capsid center. Since the shifts used in the Vol Align Tools job were calculated from a centered capsid, itâs not surprising they result in misaligned particles when applied to the particles that reconstruct the volume in the image above. In retrospect, my Remove Duplicates job in step 4 would not have worked unless the Sym Expand job moved the centers, so I should have planned for that, but I donât see any indication that recentering occurs in the CS guide page for sym expansion, nor does such a shift seem to be taken into account in the protocol I linked to earlier. They donât Remove Duplicates so could that somehow introduce the shift? It seems very unlikely to me.
Iâm going to re-measure the coordinates to use for recentering in the Vol Align Tools job from the homo recontruct only map I posted in this comment, which I expect will resolve the issues with subtraction, re extraction, etc. However I would like to know for sure the source of the âextraâ center offset Iâm observing, and how to predict or adjust its magnitude.
This is definitely not the expected behavior. Neither sym expansion nor remove duplicates should adjust offsets for something that is already correctly aligned with respect to the symmetry axes, so Iâm not exactly sure what is going on here
Yes, I have started over from step 1 of my protocol above, extracting new particles (with a larger box size) and running a homo refine of the entire capsid, and so on.
In this case, we see the expected behavior of Remove Duplicates completely reversing the sym expansion, since all the expanded particles have the same center.
I also noticed that CS keeps popping up a notification: âclearing intermediate results for P323 J249â, which is the homo refine job from which sym expansion proceeds. So, while that job is currently centered according to I symmetry, it may be that it at some point was centered near a vertex, and these results keep getting sent on as output from the job, instead of being cleared with the rest.
So, this problem can probably be chalked up to âuser errorâ, though if youâd like me to upload data etc. for further troubleshooting, Iâd be happy to help.
Thanks for this report. Based on the steps described above, what you are observing is unexpected â I have some questions to help us troubleshoot.
The image of the reconstructed density you showed after step 4 does not look as expected based on the described workflow. Sym expansion (step 3) for a point-group symmetry doesnât modify shifts, and since the capsid was refined in I symmetry from the previous step, weâd expect the reconstructed result after step 3 to be identical to that of step 2. Similarly, after step 4, remove duplicates should keep 1 of every 60 particles, completely reversing the expansion, as you described in the latest comment. At this point, the volume should still be centered. Is it possible that the outputs of a Volume Alignment Tools job was accidentally used somewhere in the first four steps?
I believe the steps you outlined are correct, however step 4 (remove duplicates) will not do as you stated. Omitting this step, and performing the final local refinement (step 10) in C1, with tighter pose search extents and priors, is another approach.
After starting over from step 1 and re-measuring the coordinates, do you notice that now the volume alignment tools, particle subtraction, or local refine results are still incorrectly centered?
Is it possible that the outputs of a Volume Alignment Tools job was accidentally used somewhere in the first four steps?
Yes, my best guess as to what happened was I used the output of a Volume Alignment Tools (VAT) job as input to a Homogeneous Refine (HR) job, and the output of that homo refine job as the input to Sym Expand, and that to Remove Dupes. Then later, I cleared the HR that had been based on VAT, and restarted it using I sym and centered particles. I restarted the Sym Expand and Remove Dupes jobs that were input from the HR job, but, for unknown reasons, the SE job continued to pull results from the original run of the HR that used VAT input, results which should have been cleared, instead of the centered, I sym results.
If thatâs the case, the Sym Expand and Remove Dupes jobs are functioning as expected, and the unexpected behavior comes from some problem with clearing intermediate results followed by clearing the job entirely.
After starting over from step 1 and re-measuring the coordinates, do you notice that now the volume alignment tools, particle subtraction, or local refine results are still incorrectly centered?
I havenât tried VAT on the re-extracted particles. I expect the centering would be correct with a âcleanâ sequence of jobs.
My current plan is to create a new HR job for step 2, then insert steps 2.5 & 2.6 into the protocol; VAT jobs that shifts the center by ~10 angstroms along the 5-fold axis that passes through my selected vertex, 2 jobs are needed for both the focus and subtraction masks. This small adjustment lets the subtraction step work for the whole capsid. (in the original protocol some of the capsid density was shifted out of the volume boundary. Re-extraction with a larger box size was to avoid this) After subtraction is done, another VAT job to center the particles etc. on the center of mass of the unsubtracted vertex should work.
I did run into further difficulties, somewhat related:
I tried to re-extract a subset of particles recentered to align shifts to the unique vertex, but the extracted particles are all centered near the center of the capsid.
The particles input to the Extract From Micrographs job have gone through the following steps:
HR with sym I
sym expand I
VAT - shift ~ 30A along one 5-fold axis
remove duplicates (4/5 of particles eliminated)
Subtract signal
VAT - shift and rotate so vertex is centered with C5 sym
downsample - crop & recenter using aligned shifts
3D classification to select 1/12 of particles with the portal, tail, etc.
sym expand C5
3D classification with sym C1, select class showing 12-fold symmetry of portal and 5-fold symmetry of capsid, with original particle count (ie, fully reversed the most recent sym expansion)
local refinement C1
Extract From Micrographs, recenter using aligned shifts
Iâve bolded the steps that I understand to adjust the particle centers and/or shifts. Skipping step 11 and extracting the 3D class that would be input to 11 gives the same result. Extracting after step 8 also gives similar results, particles centered at the positions in step 3.
Iâm aware of the issues around recentering during downsampling:
but my expectation is that the shifts applied after downsampling should still be recorded and contribute to the centering of extracted particles. Is that not correct?
Hi @miwoodso, sorry youâre still having trouble with this. If you donât mind, could you check the following things so we can try to figure out at which step this process is failiing?
After step 2 above, perform a Homogeneous Reconstruction with C1 symmetry. Does this map look identical/nearly identical to the map from step 1?
After steps 3 and 4 above, perform a Homogeneous Reconstruction with C1 symmetry. Do both of these maps display the expected shift?
After step 5 above, perform a Homogeneous Reconstruction of the subtracted particles. Does the map have the correct regions subtracted, and is it in the expected position?
After steps 6 and 7 above, perform Homogeneous Reconstructions. Is the expected vertex in the expected position in both maps? Does the step 7 map have the expected box size and shape?
After step 8, perform Homogeneous Reconstructions of each class (you can use Heterogeneous Reconstruction Only). Do all of these reconstructed maps have the expected size, shape, and position (with some differences from the classification, of course).
After step 9, perform a Homogeneous Reconstruct job in C1 symmetry. Does this map look the same as a C5 map of the particles used as an input to Symmetry Expansion?
After step 10, repeat the Heterogeneous Reconstruct Only process described for step 8. Do these maps look as expected?
