About pixel size in importing .eer format movies

Import and Export
1

Hello, everyone! I’am now doing data processing using datas collected in .eer format and I’am curious about the “pixel size” in importing movies job. The raw pixel size is 0.57 but it seems to be 0.285 when movies were imported into cryoSparc. Can anyone tell the difference?

Below are the parameters entered into cryoSparc when importing movies:

Screen Shot 2023-01-06 at 18.59.21
Screen Shot 2023-01-06 at 18.59.211084×392 31.7 KB

Below are the parameters after movies imported into cryoSparc

Screen Shot 2023-01-06 at 18.59.11
Screen Shot 2023-01-06 at 18.59.11782×342 35.1 KB

2

Hi,

Were the movies imported with an EER upsampling factor of 2? If so, this results in 2x2 super-resolution frames (i.e. 8192 x 8192 at half the pixel size), which may be where the discrepancy stems from.

Cheers,
Yang

3

Thanks leetleyang! Yes, the upsampling factor is 2 and should I change it to 1 or change the pixel size to 2x raw pixel size according to your explanation?

4

Hi,

To clarify, for EER data, cryoSPARC expects the input size to reflect the physical pixel size, prior to any upsampling. The pixel size you’ve provided is correct.

Upsampling is an optional treatment whereby the experimentalist may choose to make use of sub-pixel precision–up to 4x4–to sample beyond the Nyquist limit. Whether the sample or the dataset’s signal-to-noise at that precision justifies such treatment is a judgement call. Upsampling by 2 doubles the pixel count in each dimension, hence the halving of the resultant pixel size. That is consistent with what you see.

If you wish to work with movies that are at a pixel size of 0.57Å, then import should be performed with an upsampling factor of 1. I would argue that this should be the default in cryoSPARC instead of 2. However, if you wish to continue working at 8K x 8K scale, then bear in mind that anything expressed in terms of pixel size down the line, e.g. particle/box sizes, should be calculated with the super-resolution pixel size in mind (i.e. doubled in terms of pixels).

As an aside, you could probably also sample your EER data temporally more finely. A fluence per frame closer to 0.8-1 e/Å^2/frame may be more optimal, depending on how far your data goes.

Cheers,
Yang

1 Like
5

Thanks Yang! Indeed, there are always errors saying “out of memory” with the upsampling factor of 2. I have changed it to 1 and restarted motion correction job. Hopefully, it will accelerate the processing speed. Thanks again.