Hi all,

I need to bin the particles due to GPU memory limitations (even in final refinements). My question is about how particle binning works. The original images have a pixel size of 0.5A. I want to know if particles binned to non-integer value makes any sense. for example, If I bin these pixels of 0.5A to 0.8A (1.6x binning) will they have more information than particles binned to 1A (2x binning)?

Thanks,

Kapil

Yes, you do not have to bin by integer multiples. So long as the resultant box size is an even number you are fine

Thanks a lot. Is there any reference I can use to understand the binning?

â€śBinningâ€ť is really a misnomer we use for historical reasons. Scaling the particles is done using the Fourier cropping technique; it is equivalent to interpolation onto a coarser grid using a sinc function. The Fourier transforms can be cropped by any number of pixels, corresponding to various non-integer scaling factors.

Binning is basically some form of pixel averaging in real space which is not anymore used in most if not all of the cryoEM processing pipelines. It would be better to use an accurate terminology like Fourier cropping or in more general terms, down-sampling.

Alpay