Hi Cryosparc team
I am trying to classify phase plate data in cryosparc . I am wondering if there is some modifications to the parameters that I have to make so that the ctf correction is done correctly . It is kinda similar to the Appoferritin case here
Processing phase plate data
but centring is further complicated by my proteins size and lack of symmetry .
Any tips would be appreciated
Generally cryosparc works well with phase plate data (for example, EMPIAR 10078 T20S can easily go to 2.5 A and 2D classes are good).
Things to try for more difficult 2D classifications in cryosparc in rough order of time/usefulness:
- Increase classification uncertainty (try 4 - 8)
- Lower maximum resolution (try 12 A)
- Try a very small number of classes (15 - 20) and a large number (300)
- Solvent clamping
- Pre-whitening and solvent clamping together
- Set force max over poses/shifts to false (marginalize)
- Randomly perturb poses
Are you having OK results with Relion? Do your phase plate data power spectra look normal (i.e. stigmatic high res Thon rings)? If you plot estimated phase shift over time (or micrograph number), does shift evolve as expected? Only few micrographs far from phase shift trend line? Is the circular mask big enough (cryosparc default is a bit aggressive).
Finally if centering is poor, after first round of 2D (perhaps with small number of classes), you can use csparc2star.py to go back into relion and re-extract with refined particle centers.
Thank you for the tips Danniel and detailed answer. I will use them to troubleshoot further