Cryosparc FSC resolution vs Relion FSC resolution


After Refinement (Het. + Hom. + NU + Local) on cryosparc I get a map which cryosparc says is at an overall resolution of ~3.7A (with local resolution of 60-70% of the protein core (membrane domain) at a ~ 3.0-3.3A)

I take the same set of particles and take it through refinement on Relion and the post-processed maps looks better - I don’t see any more high resolution feature than that in the cryosparc maps but some of the relatively lower resolution domains are clearer and several of the beta-strands are also clearer in the Relion map. Nevertheless Relion says that its map is ~4.1A (with the core at ~3.5-3.8A).

I re-imported the Relion refined particles into cryosparc and did a local refinement job where I constrained the search space (angle: 1 in steps of 0.1 and offset: 0). The map looks improved than the prior cryosparc jobs reflecting the improved aspects from Relion refinement but the map resolution is ~3.7A.

I am not sure whether the ~0.4A difference in resolution is “real” so I’m a little unsure about what is the actual resolution here and how to proceed. Any advice here?

(The FSC curves of the cryosparc/relion reflect the difference in GSFSC resolution but there is nothing that is conspicuously wrong/suspicious in either of them (at least not to my eyes))

Are you using the same mask for FSC calculation?

For post processing on Relion I generate the masks but on cryosparc I leave the mask as dynamic during the local refinement jobs. The mask generation parameters are identical on both Relion and cryosparc. As much as I can say on Chimera, the relion mask and the mask from the final iteration on cryosparc local refinement are very similar.

But no - the masks used are not identical.

when you say the masks look identical, are you comparing with the refinement mask used in cryosparc or the fsc_auto mask? Just that often this can be the source of discrepancies, because csparc auto-shrinks the mask used for resolution determination

I was visually comparing the _volume_mask_refine.mrc with the Relion mask.

I just now looked at the _volume_mask_fsc and the _volume_mask_fsc_auto from the cryosparc refinement jobs and the _mask_fsc_auto is actually very tight/shrunk.

Cryosparc displays 2 sets of FSC curves, one labeled only as FSC and the FSC-mask auto-tightening. In my case, the first set, says my map is at a resolution of 3.93 and the second set (after auto tightening) says its at 3.78. So the _mask_fsc_auto is used for the 2nd set? What mask is used for resolution of the first set?

FSC (set1 FSCs)

FSC-mask auto-tightening (set2 FSCs)

yes - I believe the fsc mask is used for the first one, and fsc_auto corresponds to the second. “Loose” corresponds to an FSC calculated with the refinement mask, which is what you were comparing with relion

@olibclarke thank you very much!