Hello,

The EM algorithm, used for the 3D refinement, calculates for each projection image a probability function over all rotations (and translations). Do I have access to these probability functions (for each image)?

Hi @TamirBendory,

The full distribution of image likelihoods over each rotation/translation is computed, but not stored due to its large size and thus can’t be accessed. We do store the optimal image pose (`alignments3D/pose`

and translation `alignments3D/shift`

which is in units of `alignments3D/psize_A`

). We also store the negative log likelihood of the image given the model, evaluated at the optimal pose, and this is stored in `alignments3D/error`

.

Best,

Michael

Thanks for the prompt response! Just to clarify, when you say “optimal image pose,” do you mean the maximal value of the probability function over the rotations/translations?

Yes, that’s correct. It is the value of the image likelihood, maximized over all poses in SO(3) and all shifts in a neighbourhood around the origin. You may see it written in the CryoSPARC Nature Methods publication in the supplementary materials under equation 10. This quantity is maximized over variables (r,t).

Best,

Michael