I know about the features “Maximum resolution (A)” and “Maximum alignment res (A)” during 2D classification that control the alignment resolution itself as well as the size of output images.
- I still dont really understand what both parameters really do
- Is this the only way to generate larger output images?
- How can I save the class averages with the correct scale (for now I always save the png, check its size, adjust it to the extract size and check the scale
Did you already come across the parameter descriptions for this job (under Inputs and Parameters)
and have any specific follow-up questions regarding these descriptions?
How to you wish to use those larger images?
Thanks. I’d like to look at them in original resolution, also I often use class averages and compare them to mics or other averages I have. For that I need them in the correct pixel size
Do you mean side-by-side visual comparisons?
Yes, visual comparison, side-by-side or overlaying with other averages I have. Nothing special, but it is important to have the correct scale.
For example when I process negative stain data or compare 2D cryo classes with stain averages
Right now the only way to obtain 2D classes that are reconstructed at an arbitrary pixel size is by setting the Maximum Resolution parameter. To guarantee that it will reconstruct at the same size as the data, you can set this to a number lower than the nyquist resolution of the data. E.g. if your raw data has a pixel size of 1Å, set this to anything lower than the number 2 (Å). The desired class images are in stored in
mrc file format, in the job directory, which we recommend using instead of the
png file. (We can’t guarantee the number of pixels in the
png is the same as that which is used internally by the job). For reading these mrc files, displaying them, and generating your own PNGs from them, we recommend using CryoSPARC Tools – see example notebook doing this.
In our next update, there will be a dedicated job that allows one to take the outputs of a 2D classification and reconstruct the classes from the particles at any arbitrary box size. This will allow you to reconstruct 2D classes at an arbitrary pixel size without repeating the 2D classification job.