Hi,
If the particle 3D alignment parameters were imported from Relion, and I only want to perform local refinement in cryosparc, I notice the per particle scale factor is set to 1.0 for every particle.
Is it possible to turn on per particle scale factor in this case? I tried to turn on “Minimize over per-particle scale during alignment” and/or turn off “Use scales from input job during alignment” but no success.
Thanks!
Rui
Hi @zhangrui_wustl,
The main reason why “Minimize over per-particle scale during alignment” is turned off by default is that when the masked region is small (relative to the full volume), the comparison between the masked volume and the images will ignore a large portion of the actual volume that was masked out. This can cause scales to diverge over the iterations.
One workaround that allows the re-estimation of particle scales would be to first run an unmasked homogeneous or non-uniform refinement with the full set of particles, with per-particle scale minimization on. You can set the initial lowpass resolution to be fairly high, maybe 6-8 Angstroms, to ensure that the volume doesn’t “drift” from its original orientation. Then, take these particles and the consensus volume, along with your mask, and run a local refinement with per-particle scale minimization off, and using the scales from the input job.
Let me know if this helps!
Michael
Hi, Michael,
Thank you so much for the quick reply!
My mask actually covers most of the protein complex, so signal is not a problem here.
Is it even possible to turn on re-estimation of particle scales in local refinement?
I don’t want to re-run global search because my initial imported 3D alignment parameters have special meaning for filamentous object and I don’t want them to drift away.
Thanks!
Rui
Hi @zhangrui_wustl,
Just checking the code, you’re right – the scale minimization in local refinement parameter is not respected under the hood, and unfortunately the job will not re-estimate scales even with this set. This is definitely a bug!
We’re working on making local refinement more robust and this will be fixed in an future version. Unfortunately right now, I don’t believe there is a way for you to estimate scales without also re-aligning all of your particles. Apologies for this; I will add this to our tracker.
As a side note, we are also deploying a BETA version of helical processing to cryoSPARC in the next release, which may be of interest for you in processing a filamentous protein.
Best,
Michael
Hi, Michael,
No problem, I’ll try the global refinement to see what happens.
And it is super exciting to hear cryoSPARC is about to release a new version with helical processing! I would be more than happy to provide data for testing. Most we have are microtubule singlets or doublets, which are not so helical 
Rui