Preparing particles for ab initio

Hi all,

I have a pretty simple question. I ran some initial 2D classifications in Relion and imported my particles to cryosparc for ab initio reconstruction. I notice during import that particle orientations are ignored since I wasn’t importing particles after gold standard refinement and I do not have a random subset assignment for each particle.

I am wondering, since I did a large offset search in relion during 2D classification to find the center of my particles as I was not confident particle picking picked the center of every particle, would it be helpful to do a 2D classification in cryosparc to refine particle offsets before going to ab initio? Or does ab initio not take any particle orientation/offset values anyway?

Thanks in advance!

Initial orientations and offsets are not used in ab initio. They are only used in local refinement, 3D-VA and 3D classification.

1 Like

…though I’ve always wondered why this is the case. Of course there are no angular priors, but there are STRONG classification priors. Wouldn’t the job be significantly faster and potentially more accurate if automatically performed on a subset of each 2D class and then applied to all others in the 2D class? …in some elegant way that accounts for uncertainty etc. Why do we waste all the info from 2D classification, making the job only a “is particle/isn’t particle” filter with a largely unused metric for “probably good quality set of particles”?

1 Like