Question about solving peripheral domain


We are using cryoSPARC v3.3.2 .
I have a structure of a homodimeric complex (C2 symmetry), after Nu-refinement resolution is around 3A but only for the core domain. A peripheral domain is quite flexible, and I would like to do a local refinement for this part but exploit the symmetry of the particle. However, I am not sure how to do this correctly in cryoSPARC.

Below is what I did so far:


  2. then I followed this case study: Case Study: Yeast U4/U6.U5 tri-snRNP - CryoSPARC Guide

a) I used Chimera to make a mask for one peripheral domain. Here, I do not know whether I should have made a mask of both peripheral domains of this homodimer, or I only need to make a mask for one of them.
b) Then I prepared a mask of the rest of the dimeric structure, which includes the peripheral domain of the second monomer. Then IMPORT VOLUME: type is mask, and afterwards VOL TOOL, with option: type of input volume: mask, type of output volume: mask, threshold:0.05, Dilation redius:5.
c) Then I did PARTICLE SUBTRACTION with option: particles from symmetry expansion, with the mask of the rest of the structure, and the NU-refined structure volume low pass filtered to 6A.

d) finally LOCAL REFINEMENT (see settings in attached picture), but the map resulting from this did not look very promising.

Is this workflow correct or did I get it wrong at some point?

Many thanks for your help and looking forward to your reply!
Best wishes,

Hi @QChong,

This seems more or less fine, although I would recommend using the new version of local refinement (here you are using the legacy version). I would also try the same workflow but without particle subtraction (it does not always help). In (a), you only need a mask covering one domain after symmetry expansion, not both.

The search parameters for local refinement can have a significant effect upon the outcome.


Hi Oli,

Thanks a lot for this replying! : )

I tried what you suggested by using this new local refinement with import particles stacks from symmetry expansion (without signal subtraction).

One thing I forgot to mention in last message is that the head domain is only 50KDa and flexible (from 2D classification, I could see a blurry round shape clearly which is as same as a prediction from alphafold model).

I tried two different masks:

a) a mask only for the head domain (around 50KDa), and then do local refinement(NEW).

I got 5.58A from this, but I could not see secondary structure(it has spike, discontinuous density, because of overfitting?). My confusion is whether this is because of the head domain is too small or something went wrong with my parameter for alignment. Attached is a screen shot of local refinement, parameter which did not show here is as same as default.


b) a mask contains more than head domain(it also contains part of the core domain, totally around 90KDa), then do local refinement(NEW)

I got 4.41A from this, I could see helix but with very low resolution.

So may I ask what I could do for improvement?( I also find similar issue from previous answer from Cryosparc : About local refinement, but I am still not sure about how to change this rotation/shift, or should I make tighter mask? For example, my dilation radius is 5, soft padding is 18, coming from > 5*3.3/1.26 )

Looking forward for your reply and thanks a lot!

Best wishes,


Hi @QChong,

50kDa, if the initial density is poor, is going to be tricky to improve by local refinement (though not impossible - really depends just how flexible it is). The overfitting you notice can sometimes be reduced by the use of gaussian restraints ("Use pose/shift Gaussian prior during alignment). I think that would be worth a try, using either the larger or smaller mask.

The other thing that may be worth a try is classification without alignment - to treat this as a classification rather than a refinement problem.

Another way is to run 3D-variability analysis, and then cluster on a mode corresponding to the appearance/disappearance of the mobile domain, e.g. as described here:

Hope that helps!


1 Like

too much soft padding, right? getting the PERFECT mask for this is crucial and non-intuitive. I would try several masks generated from small changes in the regions of interest and small changes in the soft edge/dilation, dynamic vs static etc etc. along with changing values for shifts/angles. basically you’re on the right track, it may never work, but there MAY be a miracle combo of these parameters. And everything Oli said absolutely. In my experience you want the mask to completely exclude the neighboring rigid domain, or else this will dominate the alignment as a single density already well aligned in the corner of the box. I suppose the fulcrum could be modified to circumvent this but have not tried.

1 Like