Inaccurate apoferritin picking

Hi experts, I’m trying to pick apoferritin particles (~120 angstrom), directly use blob picker can not accurately pick them so I tried

Curate exposures - manual picker - 2D class - Select 2D - template picker - extract micrographs - 2D class - Select 2D

Then I use the new 2D class to run template picker again but this time the extracted micrographs are less accurate than the old 2D that is generated from my manual pick. I’m expecting the 2D class will be improved each iteration.

My question is how to accurately select apoferritin and generate high resolution 2D class, thanks a lot!

I find the ring blob with minimum of 90Å and maximum of 110Å perfectly centres apoferritin picks.

1 Like

Hi @rbs_sci , thanks a lot for your kind help!

I tried blob picker with ring blob from 90~110 Å, but a lot of pick at empty area. Here is my setting of blob picker, could you please take a look?

Here is one micrograph with and with out rings:

What thresholds are you using to set those picks? How many picks are made straight from the blob picker? If that is the raw output with no thresholding, something else is wrong.

Here’s what one of my micrographs looks like with apoF circled:

And without the picking circles:

Lowpass filter set to default (5A).

1 Like

Hi @rbs_sci , yours looks amazing! And thanks a lot for your help!

I set Lowpass filter to templates and to micrographs as 5 Å and the picking seems much better. It start to pick particles.
But still some miss picking, is there any other parameters can help this?

“What thresholds are you using to set those picks?”
Which thresholds do you mean, the power histogram from inspect picks?
I used 330-360 for power score and 0.13 for NCC Score for the picture below, but whatever I used the circle is not picking the particles.

I would double check your input pixel size - a pixel size error could cause this


Hi @olibclarke thanks for your help!

I collected data on a Arctica with K3 camera under 130k x, the data is in tif format.

I used 0.824 angstrom for import movies.

But I’ve just found K3 has a physical pixel size of 5 microns and based on the equation for effective pixel size:

50000 angstrom / 130000 = 0.384
Maybe I should use 0.384 angstrom as the raw pixel size?

When picking, leave the template and micrograph low pass at 20Å. HIgher is unsafe. I meant 5Å for visualisation in the Inspect Picks job.

As Oli says, pixel size being wrong would do this. Is this super resolution data and you’re using the pixel scale from counting mode?

If you use the measuring tool in manual picking, how large does it estimate the apoferritin particles to be?

1 Like

Hi @olibclarke, thanks for your help!!

I tried to measure the size as below, they are slight larger than 120 A (138, 132 and 146)
Can I ask why pixel size will cause off picking? Lots of particles pick in my template picker are at empty region.


Pixel size is definitely wrong. At 0.824Å/pixel, apoF should be ~145 pixels.

How large does CryoSPARC report the micrograph dimensions to be (on the Import Job card)? Counting mode K3 data should be 5,760x4,092.

1 Like

Yes it is 5,760x4,092!

OK, then check with the microscope facility that the magnification the data were collected with is correct. :slight_smile: Because that currently looks like ~1.4Å/pixel…

1 Like

Thank you so much! I know it is 130k magnification but couldnt find a chart to show how large is the pixel size. Do you know the pixel size of K3 under 130k?

That will depend on the microscope and whether there is an energy filter, how the lenses are tuned, and so forth - it’s better to check with the facility.

1 Like

Got it, thanks! I’ll try when I have the right parameters!

Is it possible that you set F-crop to 1/2 during motion correction? The calculated pixel size based on your measurements is 1.65Å, which is exactly 2x your nominal pixel size of 0.824.

What I meant re a pixel size error is that if your pixel size is off from the true value by a significant amount, then a ring template based on the “true” size of apoferritin will no longer correlate well with your particles.

So in your case, if you are measuring the particles at a nominal outer diameter of 140 Å, you might want to try a ring with a min diameter of 110-120 and a max diameter of 140-150 and see how that goes. You can always calibrate the pixel size later, after refinement.


1 Like

Hi @rbs_sci and @olibclarke , thanks a lot for your help!

I correct my raw pixel size to 0.66, the images were taken by krios under 130x, I found parameters in mdoc file.

I compared template picker and blob picker. Blob picker with ring blob and pick range 110-140, it does work better, starts to pick particle but mostly pick the edge of the particle

The template picker I select 1500 particles, the 2D was great but when I use it to template pick, the result is not good, still can’t pick the particles.

I think it may because of my data is not in a good quality? Maybe too much noise there?

Your data should be perfectly fine for picking… not sure why it is not working. What particle diameter did you set for template picking? I would go a bit larger than the actual particle - maybe 140 or 150 Å

Yes, while conc is not up there, if the pixel size is correct I’m very puzzled why this is a problem…

…although it’s interesting that picking works with both larger and smaller than actual particle size…!