Hi CryoSPARC team,
Our lab has a small protein refinement which was causing some trouble. Standard issues with small proteins now fixed by some careful adjustment of initial filtering parameters and batch sizes, however one thing has left me scratching my head…
The ab initio model is quite nicely centred in the box but when it is passed to Non-uniform Refinement with symmetry alignment enabled, it is correctly aligned to the C4 axis, but shifted significantly from the centre of the volume (about a 20% shift along the Z-axis). This causes problems during refinement.
I restarted the job a few times to see whether I could “random chance” the shift, but it’s pretty consistent. Oddly, when I use relion_align_symmetry the map appears bang in the centre of the volume consistently, calculated shifts are consistent across a dozen or so runs although the angles change to face it one way or the other, and vary rotation around Z.
Right now it’s an issue with a workaround, I’m just a bit puzzled why the Z-axis centring is so consistently “off”…
…
I need to add thanks for the addition of the easy upload facility in CS4.5, though; it’s made life a lot easier for cases like this (and custom focussed masks too!)
I wonder if it’s an issue with the automatic resampling… have you tried resampling the ab initio volume to the full box size prior to providing it to NU-refine?
Good thought - set up a test before I went to sleep last night. 4x bin particles (to initial model box size) has same result as 2x bin, so I don’t think it’s that unless the box is (further) downsampled during alignment/recentring at the start then scaled back up.
Now at 1.9A with no CTF refinement or RBMC, so going to wrap this processing up before I experiment further. If I can’t break 1.5A I’m going to be quite disappointed. 
1 Like
Hi all,
Thanks for reporting. The symmetry alignment procedure doesn’t include any bias to prefer keeping the structure centered along the z-axis, and since the origin of a Cn symmetric structure can exist anywhere along that axis, it’s possible for it to shift arbitrarily far away. We’ve made a task to encourage symmetry alignment to keep the structure’s center of mass near the origin of the box.
In the meantime it may be possible to overcome this by launching a subsequent volume alignment tools job with the volume + a mask generated around it with “Re-center to mask center of mass”
Best,
Michael
2 Likes
Looking forward to it, @mmclean, thanks! 
I didn’t realize this was happening, but have now noticed the same - would be definitely useful to fix, as if users don’t realize and then re-extract their particles using the shifts from refinement, the particles themselves will now be off-center.