It would be very nice to have an option for post-acquisition super-res during motion correction as discribed in https://www.biorxiv.org/content/10.1101/2023.06.09.544325v1
Very straightforward to implement, but useful for people who have already collected data at bin2 with Gatan K3.
For Cryosparc devs: Feel free to directly use this code:
The license is MIT.
Paper repo here, but Alex’s implementation has had more updates.
I didn’t see the realize there was a repo for that version. Yeah, the version I posted is an updated version of the one you posted.
BTW, found your account=)
Thanks for your input and links to the repo’s @alexjamesnoble and @rbs_sci!
PASR certainly looks useful for viruses or other large particles where lower magnifications need to be used to get a suitable number of particles. It may also theoretically help with datasets which exceed the Nyquist of the collection, but since that is typically ~1.5 Å, we expect that may be a less common use case. Are there other potential applications of PASR that we’re missing?
It may help for any sample which is extremely heterogeneous, as acquiring at higher magnification is no guarantee of higher resolution… particles acquired increases (per micrograph/unit time) which in turn increases the likelihood of reaching higher resolutions, while not wasting storage space if the (lower) physical Nyquist is not reached.
Since the size of cryo-EM data isn’t going to get smaller any time soon, optimal use of storage is becoming increasingly important.
For EER data and regular sized particles, PASR probably doesn’t make much sense since sampling on the 8k grid is available natively. But when collecting data on K3, it should be possible to use at least one step lower magnification, since a usage of PASR&bin2 does not require longer exposures and larger file sizes to be recorded on the K3 PC compared to super-resolution stacks.
But this, of course, should be tested (PASR&bin2 vs. Super-res bin1 vs. Bin2 one step higher mag).
I my case I was aming for 2Å at the pixel size of 0.874Å on bin2 data, but after Reference Based Motion Correction I got ≈1.8Å. So if I had used Falcon 4 and the EER format, I would have just rendered hits on the finer grid, which I can’t do now with the K3 bin2 data.
Thanks for the responses, we have noted this request!
All the best,