After some playing around I more-or-less partly answered my own queries here - reproducing in case they are of use to anyone else :)
> 1. I have a dataset from a symmetric protein that is asymmetrically disposed within a nanodisc. In such a case, I get better resolution during refinement by enforcing symmetry, but if I understand correctly the important thing during bayesian polishing is comparison of the projection of the reference map with the aligned particle image. In such a case would I be better using a C1 refinement to start Bayesian polishing, even if it is somewhat lower in resolution?
Starting from the symmetric refinement gave the greatest improvement (0.2 Å, 3.3—>3.1). Starting from the asymmetric refinement gave a still noticeable but smaller improvement of 0.1 Å.
> 2. I would like to start polishing from orientations obtained in another package (cryoSPARC in this case). I am assuming that the polishing algorithm is expecting a map that has been calculated in relion, but in certain cases I obtain substantially better resolution in cryoSPARC. In this case, could I just take the cryosparc orientations, calculate two half maps using relion reconstruct, and then use relion_postprocess on those half maps? Or does the refinement itself need to be performed in relion?
Tried a couple of different approaches here. The only one that worked was the following:
1. Convert cryosparc metadata file to star (using pyem)
2. Generate half-stars based on the random subset column of the star file using awk ( e.g. awk ‘$16==1{print}{}’ csparc.star > half1.star)
3. Use relion_reconstruct to generate half maps
4. Run relion_postprocess on the half maps (using the refinement mask from cryosparc, or another similarly generous mask)
5. Run bayesian polishing, and then put the shiny.star back into cryosparc.
What didn’t work:
1. Using cryosparc half maps in step 4 - postprocessing looks fine, but resulting particle tracks are noisy and clearly wrong
2. Using a tighter mask (e.g. the fsc_auto mask from cryosparc) for postprocessing - this gives much better resolution during the postprocessing step, but much worse results for polishing.
Cheers
Oli
########################################################################
To unsubscribe from the CCPEM list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCPEM&A=1
|