Can someone tell me the best resolution of the EMDB-10164 dataset with M

149 views
Skip to first unread message

yuqi zhang

unread,
May 18, 2021, 3:59:42 AM5/18/21
to Warp
Can someone tell me the best resolution of the EMDB-10164 dataset with M?

I used the series 1-20. I used the EMD4045 to template matching. And I matched near 2500 particles and did refinement in Relion 3.0. After M refine,  I only got 10A.

I wanted to know whether I have to do or learn more about the workflow.

Alister Burt

unread,
May 18, 2021, 4:51:28 AM5/18/21
to yuqi zhang, Warp
Hi Yuqi,

2500 particles per tilt-series or 2500 particles from 20 tomograms? If the latter, that’s very few particles. For reference, we used 19810 particles from 5 tilt series to get a 3.4A reconstruction from the same dataset (full walkthrough https://teamtomo.org/walkthroughs/EMPIAR-10164/introduction.html). I believe Dimitry has the same subset at 3.3 or so - the full dataset has been pushed to 3.0A (as described in the M paper).

I had a similar problem at some point with refinements of this data getting ’stuck’ at 10A or so (in RELION) - at some point I processed the same data tagain, the only differences were in the ‘don’t invert tilt-angles’ on tilt-series import and subsequent defocus handedness checks, one refined to below 4A no problem whilst the other got stuck at 10A. The procedure described in the walkthrough linked above works in our hands.

Good luck!

Alister

--
You received this message because you are subscribed to the Google Groups "Warp" group.
To unsubscribe from this group and stop receiving emails from it, send an email to warp-em+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/warp-em/5cb54a4e-6ba2-4dd7-a0b7-54c8aba597ddn%40googlegroups.com.

yuqi zhang

unread,
May 18, 2021, 7:55:22 AM5/18/21
to Warp
I remembered I checked the ‘don’t invert tilt-angles’ and detect the defocus handness. 

yuqi zhang

unread,
May 18, 2021, 8:25:58 AM5/18/21
to Warp
I checked   ‘don’t invert tilt-angles’  both when the import tilt series to IMOD and from IMOD. Is that OK?

Alister Burt

unread,
May 18, 2021, 8:42:02 AM5/18/21
to yuqi zhang, Warp
During stack creation this parameter doesn’t affect anything (tilt angles in .rawtlt are written out the same), only during tilt-series import (where the tomostar files and tilt-series XML files are created) is this checkbox taken into account.

According to https://teamtomo.org/walkthroughs/EMPIAR-10164/2d-preprocessing.html#stack-generation we did not check the ‘don’t invert tilt angles’ box when we processed successfully. The alternative should also work, and I have heard of others processing successfully using that checkbox, it just didn’t work in my hands when I had that option checked and rather than investigating more thoroughly I accepted a working procedure once I had one.

Cheers,

Alister

yuqi zhang

unread,
May 18, 2021, 8:43:49 AM5/18/21
to Warp
OK. I will follow the tutorial link with dynamo. And check about this.

Alister Burt

unread,
May 18, 2021, 8:50:17 AM5/18/21
to yuqi zhang, Warp
If you already have .xf files etc you don’t need to do anything in Dynamo - feel free to only follow the Warp/M specific parts!

If you change the ‘don’t invert tilt angles’ checkbox you will have to repack particles, tomograms will be flipped in Z.

Cheers,

Alister

yuqi zhang

unread,
May 20, 2021, 9:44:16 AM5/20/21
to Warp
Now, I finished the guide link. I picked 20402 particles in 5 tomograms, and got a 7A resolution map. I used template matching with emd4045 rather than using Dynamo to generate a template. Maybe I will test relion classification in the future. I cannot get a near 3A map right now.

Alister Burt

unread,
May 20, 2021, 2:41:08 PM5/20/21
to yuqi zhang, Warp
Good job! Was 7A after relion or did you also run M?

Possible issues:
- particle set contains junk
- initial tilt-series alignments weren’t great

Making sure you take the time to isolate a clean set of particles and that their orientations make sense with respect to their supporting geometry is important.

Cheers,

Alister

yuqi zhang

unread,
May 20, 2021, 2:44:27 PM5/20/21
to Warp
8.2A After relion. And 7.7A after M.
It seemed like my computers cannot run the 3D classification with 20000 particles due to GPU memory limit.

yuqi zhang

unread,
May 20, 2021, 2:46:00 PM5/20/21
to Warp
And another question is: the particle box in relion was 128*128*128 pixels. After M refine, it became 200*200*200.
Reply all
Reply to author
Forward
0 new messages