Bin micrographs

Hi,
After ctf correction, how can I bin the micrographs. I need to do it as my PC configuration will not allow boxing out particles with a box size more than 1024X1024 pixels.

How can I get CTF corrected micrographs and binned micrographs (at different stages of processing).

If I use a different package to bin the micrographs and try to introduce the binned micrographs in the cryosparc pipeline, it complains about not finding some files like ‘json…’. Any way it can be done ?

Regards
Andy

Hi Andy,

Do you really need to bin the micrographs themselves, or just the particles? The latter is probably preferable, as it will allow you to adjust the degree of binning later. You can specify this during particle extraction (“Fourier crop to box size”, or similar).

Otherwise, you can also Fourier crop output micrographs during motion correction. This is often done when processing super-resolution stacks.

Cheers
Oli

HI Oli
Thanks for the reply. Binning the particles will be ok but the problem is that the max box size allowed by the program (and based on the configuration of the machine I am using) does not encompass the whole particle. So binning it later will not suffice UNLESS the box size remains the same only the image gets binned. If thiol can be done then let me know how- it will save a lot of scripting labor.

For Fourier crop during motion correction - how it effects the CTF estimation at high resolution on the later stages of processing ?

Regards
Andy

Hi Andy, what exactly is happening when you use a larger box size? When you say it is failing, what do you mean?

I am suggesting using a larger extraction box size, then Fourier cropping to whatever size you desire. This does not change the size of the region that gets extracted, only the spatial sampling of the resulting particles.

Cheers
Oli

HI OLi,

The max value of the box size in pix cannot go over 1024 when I am using manual picks to generate template. Following to that how particle extraction will work as in the previous step I am unable to encompass the whole particle in the box.

Best

Andy

Hi Andy - I think there is some confusion here - are you talking about the slider in manual picker? This is only for visualization - it has no relation to the value you enter in the extraction job (either CPU or GPU based). Just connect an extraction job to the output of manual picker and you should be fine.

Hi Oli,
So the idea (old school) is to get some particles picked, generate 2D class averages and select some good ones to use as a template. When I try to box the particles for the above job using box size 1200 x 1200 pix, I got an error saying that I can use only 1024 x 1024 pix boxes. So I took the aligned micro graphs, binned it with a different program and then re-imported imported then into cryosparc (a good deal of time went in for the import).

So my initial question was and is - can I bin the micrographs before picking the particles, manually.

Hope this explains the situation.

Best Andy

right - but how are you trying to box the particles. Are you using the Extract Micrographs job type? Because I am pretty sure this can handle >1024px boxes. In which case just do that, and choose the appropriate Fourier cropping.

Not sure if we are on the same page- i didn’t even proceed to extract the particles as my box cannot encompass the whole particle (but only a part of it). So I binned the micrographs separately and then boxed it.

Do you indicate that even with a small box size to assign the coordinates during particle boxing , while extracting them - one can use a bigger box size and then do Fourier cropping ?

I think you are getting confused by the slider in manual picker - this only affects visualization.

You can extract the particles in whatever box size you want, using the extraction job.

1 Like

That is what I was trying to clarify. Thanks

I have figured out the issue and it is solved. However I now ran into another problem-

After extracting the particles (> 1 million) and starting 2D classification- I encounter “cache full …”. I tried to empty space on the cache (1TB SSD) without much success. Is there a way to carry out the job (without adding another SSD).

I split the whole data set (extracted particles in 6 groups (in different folders) and tried to load one set at a time for the 2d classification but it failed.

Let me what can be a good solution.

Best Andy

Hi Andy,

1M particles, if sufficiently binned, should fit into 1TB scratch just fine - I would recommend binning to 4Å/pix initially, or 8Å/pix if your stack is still too big to fit on the SSD.

Cheers
Oli

Hi Oli,

I am placing my next set of questions here that remain unanswered for more than a week, if you can respond to these.

1.While the class averages are well centered, a major population of the auto picked is completely off centered. How can we center the particles in cryo SPARC those going in the reconstruction process .

  1. I binned the autoboxed particles to carry out the 2D class averaging. After the selection of particles from the correct class averages- How can I use the un-binned forms of the selected particles for reconstructions.

  2. how can we set up finer angular steps and shifts in orientation search and refine processes in the reconstruction process.

  3. In the reconstruction process (refine) how can we regulate the
    particle selection based on Figure of Merit or any other criterion to make selection of the best particles from the selected band, suitable for the highest resolution of the density map.

Thanks
Andy

Hi Andy,

  1. If the particles are off centered, but the 2D classes are well-centered, re-extracting with recentering (which is on by default) should recenter them.

  2. You can do this by replacing the particles blob in the low level results with the one from the “full scale” particles - see (Initial resolution improvement, then deterioration in Non-Uniform refinement - #9 by wtempel) for an explanation of the process in another context.

  3. In global refinement, this is handled automatically, and I do not believe there is a way to “force” finer angular samplings. In local refinement there is an option to choose the maximum angular sampling (by default 0.2deg).

  4. There is no way to select based particles based on figure of merit (I guess you are thinking like Pmax in relion) in cryoSPARC, to my knowledge. You can select particles based on probability of belonging to a particular class, after 2D or 3D classification, using the Class Probability Filter job type.

Hope that helps

Cheers
Oli

Hi Oli,

Few more questions,

  1. How do I introduce an already computed moderate resolution density map as a starting model in the workflow of cryosparc.

I have the particles boxed out, class averaged, classes selected and particles extracted for reconstruction. At this point instead of using Ab-Initio reconstruction I want to use the pre-computed map as the starting mode and continue with the refinement.

How do I get the program (cryosparc) to read this precomputed map into its pipeline.

  1. How can I get the extracted particles (in mrc format)

Thanks.

Andy

  1. Import 3D Volumes

  2. Not sure what you mean - are you asking how to export to read into relion? Or another program? It will depend on what you want to do with them (in terms of how best to export)

Hi Oli,

Thanks for the 1st reply

for point no. 2, yes, - how to read the particles in another program - like say BSOFT or EMAN2 etc. I want to do some pre-processing for starting the model refinement or generate a starting model using BSOFT or EMAN 2. Can I get the particle images (stacks) from each micrograph (after frame alignment and dose correction).

Also can I get particles ctf corrected before the reconstruction process ?

Particle stacks will be located in the extract sub directory of the extraction job.

No, I believe particles are CTF corrected on the fly, so there are no stacks of CTF corrected particles.

HI Oli,

Continuing the thread here-

While carrying out class averaging for particles I encountered " illegal memory access was encountered". I do not find a solution to overcome this. Additionally this happened when I am trying to add another set of data to the existing one (which already provided a 3.4 Å resolution map).

Any feedback and a fix to the problem will be helpful.

Best
Andy