Help with merging particles from 2 datasets

Dear cryoSPARC community,

I am wondering if I am doing something fundamentally wrong when I am merging particles from 2 datasets.

I have two datasets (dataset A ~109K and B 280K particles) that were collected at the same pixel size. I’ve processed the data individually and I got maps at ~4.5Å and ~3.5Å for each dataset.

I re-extracted particles at the same box size and tried to combine them by simply dragging particles from both data sets as input to 2D classification.

Based on the output (2D class representative picture attached) it doesn’t look like the particles are being merged – The contrast is very different. Similar results was observed when I run an ab initio with 2 classes.

Any thoughts, ideas, on the matter would be much appreciated

Hi @gptomaleri,

Were the two datasets collected during different collection sessions? If you haven’t already, you might want to turn on Min over scale after first iteration in 2D Classification – this will re-estimate the contrast (i.e. multiplicative scale factor) of each particle, which can help when the sample and ice characteristics vary between the datasets. Let me know if that helps!

Best,
Michael

2 Likes

Hi @mmclean,

Thanks for getting back to me so quickly.
Yeah, we have collected them 3 months apart at the same microscope. That means that our dataset A was processed in an older version of the CryoSPARC, I doubt that this is a problem, right? The only difference between the datasets was the spot size…

Now, I’ve turned on Min over scale after first iteration in 2D Classification but there is an error message:

I tried to turn it on for datasets A and B individually and the same problem occurred.

Any thoughts?
Best,
Gio

Hi @gptomaleri,

Thanks for the additional info. Do you know the version of cryoSPARC that dataset A was initially processed with? I assume this means that motion correction, CTF estimation, and particle picking/extraction was all done with the older version?

Regarding the final error, are you currently on Patch 210615 or later? We resolved a bug in that patch that may be responsible for the ValueError – do let us know if updating to patch 210615 (or the following patch 210622) doesn’t resolve it.

Best,
Michael

1 Like

Hi @mmclean,

Yeah, we use V2.15.0 for motion correction, CTF estimation, and particle picking of dataset A. Do you think that could be the problem?

We recently update (06/28/2021) to V3.2.0 and still got the error.

Any help?
Best,
Gio

Hi,
I encounter a similar problem processing two datasets collected months apart on the same microscope with same parameters. In my case one of the samples had substantially thinner ice.

I found out the reason for the mismatch in my case. It was because for one of the datasets some initial frames were not included in the motion correction while the other one included all.

Dear all,

I have two datasets of the same protein sample collected on two different dates 13 days apart with the same microscope and settings (apix, dose).

I processed the data on CryoSPARCv4.2.1 in two separate workspaces from motioncorr to 3d refinement before trying to merge the two sets of particles. Both sets of particles (same box size) were exported and reimported to match the pixel size (prior error mismatch in pixel size up to 4 decimal places).

However the 2D classes seem to show opposite data sign (screenshot 1 below). I tried “minimize over per particle scale” during 3D refinement (screenshot 2 below) and “Min over scale after first iteration” for 2D classification, both approaches did not seem to help in correcting the scale difference (assuming if this is the root issue).

I am very puzzled and hope to seek any insights please on how to fix this and what I had done wrongly.

image

image

How were the two sets extracted? Some jobs (I think local motion correction?) generate particles with a different sign. Usually this is not a problem (it is recognized automatically) but it could be if both are present in a single job.

Yes, both sets of particles were exported after local motion correction. Curious why would the data sign be changed in this case?

Will export the particles before local motion and try again.

Thanks Oli!