I am fairly new to cryoEM and CryoSPARC. I am working on a membrane protein with a substantial transmembrane domain and am having trouble seeing resolving transmembrane helices and ectodomain in 2D and 3D classes. I have experimented with many 2D classification parameters including mask diameter (to remove influence of nearby particles - my micrographs were quite crowded in places), initial uncertainty factor, force maxing over poses on/off, increasing number of EM interations, and Minimum alignment res to 40 (to remove low res information from nanodisc).
After all of this, it still seems that the default settings produce the best classes:
As you can see, there are multiple classes with an apparent ectodomain, but some with empty nanodiscs, and some with noise that don’t align well. Apart from problems with the data itself, I think there is an alignment issue in that the classes are aligning to the nanodiscs rather than the protein inside, which may move relative to the nanodisc, hence why we don’t see it.
First thing, increase your box size by at least 25%. Preferably more, although not more than 50%.
Increase online EM iterations (probably 100 or so) and full passes (say, 3) and classification uncertainty (I like 5 most of the time). Select all classes which are not ice blobs, etc. Repeat 2D a couple more times. Possibly pick all sides, then all tilts and tops into two separate runs and 2D classify them separately, then re-combine for ab initio. When moving to ab initio, target a moderately large number of classes to see whether you can tease out the nanodisc-only particles to be discarded. Play with the start and target resolution to see what you can get.
Or, clean the stack obsessively until you’ve got a small number (10,000<n<30,000) of particles to feed to an HR-HAIR strategy to try to get a nice model, then feed back the cleaned stack into a heterogeneous refinement with the nice model and some junk targets.
If you’re playing with highpass filtering to try to reduce the impact of the nanodisc, you’ll need to increase that a lot over 40 Ang. I wouldn’t usually recommend doing this unless your protein is small and completely embedded.
