I should first say that I don’t really understand what the sign of the Ewald sphere correction stems from. Is it microscope specific? Anecdotally, several data sets collected on the same scope with similar settings all produce data which require the same sign of the correction. If it is not scope-specific, is it a property of a data set (if you process the same data set multiple time, will it always benefit from the same sign)? Or somehow of the reconstruction, e.g. depending on the starting structure? Or is it per-micrograph specific and the ”dominant sign” will be the one yielding the best reconstruction?
My questions all stems from the question of what happens if you merge two data sets. Would you need a different sign correction for each data set? At the moment, one can only set one sign for a reconstruction job. Is that an oversight and one would want to have a different sign for each e.g. optics group?
I would first ensure to refine/reconstruct the particle stack in the correct hand.
From what I understand there is no way to predict EWS, you have to compare no EWS with EWS=pos and EWS=neg. Either way the difference is small for most complexes under 1 MDa, but every little bit adds up.
*Homogeneous Reconstruction Only can be run fast to check both the hand and EWS sign.
If you imported the .xml you could explore each optics group, but you would be giving up a lot of particles unless there were many movies on a perfect sample. I do not believe EWS on a single or sub-class of optics groups would do that much.
Our understanding is that the Ewald sign is related to the handedness of the micrograph images themselves when they are imported into CryoSPARC. By handedness, I’m referring to the number of “reflections” that the images have undergone, relative to the sample’s original orientation in physical space under the microscope:
Because an unknown number of mirroring operations may have been introduced during the data acquisition and image processing, the actual direction of curvature of the Ewald sphere is often unknown for a given data set Zivanov et al., 2018
Because of this, the ewald sign is expected to be dependent on anything affecting data collection: the camera, collection software, the file formats the movies are saved to, etc.
The only way to figure out the sign that I know of, is to take each subset of data collected independently, and reconstruct each subset with both signs, and use the sign that is better for each subset. This is possible in CS by keeping particles from one subset separate from the other during a reconstruction only job.
Unfortunately it is not possible to combine datasets with different ewald curvatures right now in a single reconstruction/refinement in CryoSPARC. It would be possible to do this, however, by taking any subsets of micrographs with the negative ewald sign (or positive), reflecting them along one axis, and then writing them back out as images, and finally importing these back into CryoSPARC (then repeating CTF estimation and everything downstream). Now these imported movies will have a flipped ewald sign compared to their original sign. If you run into a case where this is required though (i.e. that different subsets of data have different signs), let us know, as we think this is a rare situation.