I’ve typically used relion in the past and while trying out cryosparc I’m encountering issues with dynamic masking trimming off most of my initiation factors during refinements- when I run a job involving dynamic masking, the algorithm seems to heavily favour the 40S body and trim off the peripheral initiation factors. Even making big changes to the masking settings only seems to marginally improve this trimming effect. I’m guessing this is due to the extreme difference in stability between the core and the extremities, but of course the interesting factors are out on the edges
Anyone have tips for overcoming this sort of situation, or any cryosparc-specific tips for ribosomal complexes in general?
If you’ve tried decreasing the “Dynamic mask threshold” param to no avail, you can also disable dynamic masking by inputting a mask to the refinement job. This gives you more control over what exactly the mask is, and the mask won’t be altered between iterations. (Note that dynamic masking is mostly useful as a convenience, to allow refinements to get to the highest possible resolution without the necessity of manually creating masks, but in cases like this it totally makes sense to disable it!).
You can create such a mask by taking the initial volume and passing it through a “Volume Tools” job, with the “Type of output volume” param set to “mask”. Then, you can manually control the threshold and dilation/padding parameters to ensure that the output mask is as wide as you’d like, and soft enough. You can also download the mask that was created to make sure it covers the whole structure.
Thanks Michael. I’ve previously given adding my own mask a shot (without much improvement, as I think the mask was still too tight) but I’ll absolutely work on tuning the dilation and padding parameters to improve things.
Another way to run a refinement without a mask (always a good experiment in a complex or flexible structure) is to set the dynamic mask start resolution to 1 Å, which of course will never be reached.
Have you tried non-uniform refinement? I work on quite similar samples and this greatley helps (even with the dynamic mask on if you go straight from binned to full size).
non-uniform refinement gave me some pretty massive artifacts in the image when I tried it. Possibly my particles weren’t homogeneous enough at the time or something, but NU refinement seemed to perform worse for this sample in my hands.
Interesting to hear that it’s working for you though, maybe I’ll play with some parameters and give it another shot.
You’re not wrong about 40S preferential orientation but my particles are mostly fine, obviously some bias but it’s within the acceptable range in my opinion. I’ve cleaned up my dataset a bit- I’ll run it again and see if that’s made any difference.
