I’m working with a protein complex (~200 kDa) attached to nanodiscs. After 2D classification, I get particles that clearly show the nanodiscs with some additional density that I suspect could correspond to the protein complex. However, the resolution is low both for the nanodiscs and for the “protein complex” portion (see 2D classes below).
I attempted particle subtraction of the nanodiscs to focus the 2D classification on the protein density, but so far this hasn’t yielded meaningful classes (example 2D classes below).
I’d greatly appreciate any suggestions on how to move forward. Specifically:
· Has anyone had success improving resolution in similar nanodisc–protein datasets?
· Would you recommend different masking or subtraction strategies?
· Should I try more extensive cleaning of the particle set, or re-examine classification parameters (e.g., class number, circular mask size)?
· Are there alternative workflows (e.g., focused 3D classification rather than 2D after subtraction) that may be more effective here?
My protein complex is tethered to the nanodiscs by a tail with a PTM; there are no transmembrane domains. We removed empty nanodiscs through gel filtration.
Nanodiscs look to be helping to get general location of your smaller protein but hindering in getting its orientation. You have some interesting looking 2d classes where it looks to be aligning well. I’d probably try topaz picking on micrographs with a center on the small particle. To train up topaz take the current 2d classification with the nanodisc and put that into manual picker and manually recenter the pick on your protein, hopefully can see it the micrographs, do this for 50-100 images; i like to turn the particle size to 15 or so on my screen so its just a dot then right click to remove and left to add a new particle centered on your small protein. Run that manual pick through 2D classification, here run a couple jobs with different size real space windows to get one that masks out the nanodisc but keeps your protein. Use the 2d small protein classification and perhaps increase the number of iterations to even more if not converging.
For topaz i like to run the train job in parallel with the jobs having different particle count number with -20% -10% 0% +10% +20% expected particle number (# picked in manual picker) and queue up the topaz extract and look at the extracts to see how it picked.
The interesting alignments? Only 3k so not many of total.
Not sure it 3D would help as relative location of your protein to overall nanodisc may not be the same. So focusing on the small protein in 2D seems to be better approach as currently nanodisc is centered.
I think it is because of the flexibility - nanodiscs seems to be overwhelming compared to your protein so 2d classification is actually aligning nanodiscs. As T_Bird mentioned above, you can try to pick those relatively rigid particles by topaz picking.
