Hi All,
Obligatory “very new to cryo-EM, sorry for what may be a silly question”. I’ve been working with a tetrameric ion channel whose structure has been previously determined and published. This channel, as well as its other family members, has always adopted C4 symmetry.
In my dataset with this channel I’ve had significant difficulty resolving the TMD; the density for this region is quite poor despite several attempts with local refinement. My highest NU-refinement resolution in C4 was 3.15A with very weak or missing density in the TMD.
I decided to take my clean particle set of ~200K particles (out of around 960K) and run a 3D classification. I saw several interesting classes. I see my C4 (which looks like the published C4) volume come up with improved TMD density, but I also see some volumes with clear C2 symmetry and great TMD density. When I run NU-refinements on these C2 volumes I get ~3.5A resolution, but really the volumes are quite nice overall and I think I could easily build the TMD region as it’s very well defined.
I guess I don’t know whether or not to trust these results. I can make an initial, broad guess as to why I am seeing this C2 volume but I’m agonizing a bit whether this is just an artifact of my data processing?
No reason not to trust it IMO - not shocking to have C2 states of a tetrameric ion channel. How are you running the 3D classification? Entirely possible to have a mix of C2 and C4 states I would say
You can take any particle set (including just non-canonical classes) and start at the beginning no bias. Ab initio, 2D, het refine, nu-refine, 3D class etc. Build your own confidence. Also great if 3D Var confirms what you saw above.
Hi @olibclarke,
I’ll go over the whole workflow just in case I am overlooking something.
After picking (using both a template and a blob) I run multiple rounds of 2D classification. From my 2D class it appears that I have a preferred orientation issue (only see the bottoms and some sides, no top views).
Once I get to around 200-300K particles that appear to be clean I’ll run ab-initios in C4. I’ll pick the best looking volume (which has poor TMD density), extract unbin (box size of 560), and run a NU-Refinement in C4. From this NU-refinement I’ll take that particle stack and run a 3D classification job (default parameters) with 10 classes and the solvent mask from the preceding NU-refinement job.
I’ll then take the volumes that come out of 3D classification and run NU-refinement in C2 or C4.
The C2 refinements are looking very promising but I don’t want get ahead of myself… Do you think I am overlooking something?