Depositing local and global maps and coordinates

I am working on a SARS-CoV2 Spike antibody complex and obtained an overall GSFSC of 3 A using NU refinement. The interesting antibody binding interface is not well resolved in the global map, with the antibody showing up as no more than a blurred blob of density. Local refinement helped a lot here and resulted in a GSFSC of 3.6 A. After post-processing with DeepEMhancer and Resolve DM I was able to build a nice model for the interface. Unsurprisingly, maps for the antibody from local and global refinement do not superimpose perfectly.

What is the best practice when depositing data at the PDB in this case?

A) Deposit coordinates of interface parts together with local refinement map (+ post-processed maps) only?
B) Build complete spike into global map, rigid-body fit locally refined parts of model into global map and deposit along with A?
C) Something else.

The problem with B) is that the blurred density in the global map does not 100% agree with the pose of the antibody derived from the local refinement.

Thanks, Guido

I would make two separate combined EMDB/PDB depositions. Deposition one: global map, model of entire Spike with locally refined parts rigid-body docked and merged (your option B I think). Deposition two: focused map and locally refined model region. Both depositions should include the raw map, sharpened map, half-maps, and refinement mask. Note that in the new EMDB website, unlike the old one, identifying and separately downloading auxiliary maps is very easy, so don’t hesitate to include anything that might be appreciated by others in the field. IMO everything else can be generated from the 4 maps I listed.

I went back and forth how to do this for the several Spike:Ab complexes from my 2021 paper, in the end I think what I propose is the best for this case. Another alternative is to make one deposition for option B but include all the local maps / masks as auxiliaries. However, it’s more confusing to consume IMO.


Thanks a lot for that answer and especially the link to your 2021 paper that features a fantastic methods and supplement part usually lacking in journals of this caliber.

It’s a bit off topic, but can you specify the input of the 3D classification (Beta) jobs you used to separate different conformers? From figure S5 it looks like you used all non-junk particles from 2D classification together with a pdb of the apo trimer as an initial model in 3D classification. Can you elaborate how you did generate the input for 3D classification (Beta)?

@GHansen The skip-align 3D classification in cryoSPARC wasn’t released at the time. I use 3D classification in the generic sense. I think I also indicated when local vs. global alignment searches were made. So 3D classification w/ global search would be “heterogeneous refinement” in terms of cryoSPARC job types.

1 Like