Patch motion correction yield worse than full frame or Motioncor2

Hi cryoSPARC developers,

I just collected some raw images by Falcon 3EC in counting mode (total dose 50 e-/A^2, exposure time 75s, for 75 frames) controlled by SerialEM, the cryoSPARC live I used (Master running v3.3.2+220824, worker running v3.3.2+220824) for the real time processing yield bad motioncorrected images, then I tried the MotionCor2 or the Full frame motion, they work pretty nice~ Not sure what happened?
Pls see the screenshots of the motioncorrected micrograph (Top: patch motioncor in Live; Mid: full frame motioncor; Bottom: Motioncor2)



Many thanks!

Best,
Henry

@stephan @DanielAsarnow @olibclarke

Any clues for this weird thing? Thanks!

Hi Henry,

What do the motion plots look like in Patch vs rigid motion?

I’m not sure what the issue is here, although there I do notice an odd stripe on the left side of your micrograph - perhaps gain correction is not optimal? Also, have you tried running Patch Motion outside of Live (I don’t imagine it should make a difference, but just in case…)

Cheers
Oli

Thanks, Oli!

Below are the motion plots for Patch motion (outside the Live, upper two) and Full frame motion (bottom one):
屏幕截图 2022-10-02 204449


I am not a advanced user, and did not find the gain reference files for this SerialEM controlled Falcon 3EC camera ~

Thanks!
Henry

Hi,
Falcon3 does not need gain correction, images are already gain and bias corrected “during” acquisition.
Regarding your issue: in my experience patch motion gives notably better results than MC2 without exception.
Did you consider to set total electron flux in cryosparc vs flux per frame in MC?

Best,
Tarek

1 Like

These plots look very odd - more or less zero motion in X and Y for 20 frames, then rapid motion all of a sudden - this is not expected. Do you see similar plots for other micrographs? It also doesn’t look consistent with what you see for the rigid motion plot from full frame. I agree with @tarek to double check your input settings for patch motion vs full frame & MC2.

image

Thanks, Tarek!

Do you mean the total dose for CS and dose per frame in MC2? I set the total dose for CS, and dose per frame for MC2 wrapped in Relion, but MC2 wrapped in CS seems not need to fill this dose per frame parameter~

Thanks!

Thanks, Oli,

I picked two other micrographs randomly from CS Live, and also my settings for Live session, pls see the screenshots below, the CTF estimation also looks odd from those patch motioncorrected ones~

Fig.1 the one I showed before

Fig. 2-3 the other two


Fig. 4 my settings for Live

Best,

Henry

@Zheng.Henry This suggestion may warrant further investigation.
Please can you post another sample micrograph in a way similar to your original post (any single motion correction option should suffice for this purpose) and a screenshot of the Exposure Group definitions (right-hand panel under the Live session’s configuration tab).

Thanks, wtempel!
Here is another sample micrograph (top, patch motioncor; mid, full frame motioncor), and the Exposure Group definition (bottom) ~



Thanks!

@Zheng.Henry Please can you also post a screenshot of the actual micrograph.

@wtempel Here are the screenshots:
from upper to lower: Patch motion; full frame; Motioncor2; SerialEM displayer




Thanks for your quick response!

Henry

A few more questions:
Have you compared the downstream (reconstruction, refinement) performance of the workflows with the alternative motion correction approaches?
Do patch motion correction results improve with Allow variable dose enabled?
Does your microscope facility have any suggestions regarding the (apparently invariant between the movies) vertical line and corner fringe patterns?

Thank you for your insights, @wtempel !
All the micrographs of this small dataset (~700 micrograph) exhibited a similar behavior like above, from the 2D average of Live I can not get a clear view; but the MC2 with Patch CTF warpper in CS yield nice 2D averages~

The 2nd suggestion worked like a charm! I just tried this, now the motioncorrected images are comparable with mentioned above, see screenshoted below


I guess the black lines may due to the camera defects, and the corner fringe patterns are because the beam off~ I will consult our EM manager for this~

Thanks a bunch, and Have a wonderful night!

Best,
Henry