Noisy Result After Homogeneous Refine

Hi everyone, I’ve been attempting to refine two structures with homogeneous refine (new) in cryosparc (the most recent version) and I’ve been getting a lot of noise outside the structure density for the final volume for both of the structures. I was wondering if anyone else has been experiencing this? One of the models I have refined in relion to 4.1 Angstrom and I have never seen this type of noise come from the data. Is there possibly a masking strategy that anyone recommends or another solution? I would like to keep my data in cryosparc through the non uniform refinement step so a solution to this problem would be much appreciated. Thanks!

Yes, I have seen this, I think, with recent versions - is the noise structured, kind of looks like a grid? In my case this only appears in the sharpened volume, not in the unsharpened, so I just take that and sharpen outside cryosparc


PS - this is the kind of noise I am referring to:

1 Like

Hi @adtaheri, @olibclarke,

Thanks for reporting this. I believe, as @olibclarke diagnosed, this artefact only appears when we sharpen the maps for plotting during a refinement. The “grid” looking feature is due to a single high-res Fourier component having a degenerate value due to numerical precision. This doesn’t affect the actual unsharpened maps or resolution estimates. If you take the output of refinement and re-sharpen it either using the cryoSPARC sharpening tools or another software, you should not see the artefacts any more. We are investigating a fix for this - if there is any additional information you can provide that would be very helpful.


1 Like

Yes! It is a grid like view exactly like that and in chimera it appears like that too. So if you take the unsharpened map and sharpen outside of chimera do you recommend taking the unsharpened map through nonuniform refinement for better results?

Thank you for your response and I appreciate the feedback!

Another question would you recommend taking the map unsharpened through non uniform refinement or the map sharpened outside cryosparc through nonuniform refinement @apunjani

Hi @adtaheri, in general, sharpening is only meant to improve visual interpretability of the map. You should always be using unsharpened maps as input for further refinement or classification.

Also, all refinements in cryoSPARC start by low-pass filtering the input volume in order to use it as an initial model. That low-pass filter by default is 30A meaning that all the high-res structure (that would have been visually aided by sharpening) is wiped out before going on to refinement. This is very important, since the initial model should not bias the high-resolution alignments or reconstruction in the next refinement. This also means that even if you did sharpen an input map, it would not substantially affect the results of the next refinement.

1 Like