The dip and curve looks a somewhat better in your local refinement.
I would tend to think that binding some lipids wouldn’t have the same effect. That said, I don’t know what you are expecting to see, will it be a large surface that binds to the lipids? Does the protein also partially penetrate bilayers? If so then you may see some disorder in the FSC curve due to the lipids, I would expect also that they would be pretty prominent at low resolution of this were the case.
What does your orientation distribution (or cFSC plot) look like?
Looking at your 2Ds, it looks like you have a preferred orientation (plus some dissociation of the dimer), presumably due to binding the lipid monolayer in a single orientation.
perhaps try collecting with a tilt? Or decorating some liposomes or nanodiscs rather than using a lipid coating on the grid itself?
It’s a bit complicated due to the different conformations we’ve been discussing, but yes the lipid-binding surface area is along the long axis of the elongated dimer, and I do see what I believe is the lipid surface in many of my 3D volumes at this point. However, I do think you were correct that the FSC looked like that largely due to heterogeneity. When I sort the stacks more extensively, the curves look much better. The resolution appears to drop too but I suspect it was basically that low beforehand.
I’m not sure my version of cryosparc has that–I’m running 4.3.1 and won’t be able to update immediately–hoping to try to update within the month because I think several of the new tools and patches would be really helpful.
There is for sure a preferred orientation. The way I’m dealing with it at the moment is I ran a 2D classification from particles that had already been cleaned fairly well. I manually made three sets, one with all the reasonable classes preferred and rare, one with all the classes in the preferred orientation and another with all the rare views. I made a separate set with all the rare views and like three of the best preferred orientation classes.
It’s possible Rebalance 2D would have done better at this task, but I haven’t had great luck with it in the past. Maybe I’ll try it again. But I know if I could, I would be using the Rebalance Orientations job. My version just doesn’t have it.
Thankfully I do have plenty of tilted data. However a large chunk of it I’m not even touching right now because everything I reconstruct with its particles looks terrible, so I have to troubleshoot that later. I have a fair amount of data at 0/15/40 degrees that behaves more normally (and is what I’ve been working with), but it seems like the low SNR of my sample with the lipids or something else about my grids may have thrown off their hole targeting automation because a lot of the micrographs are over the mesh. That’s why I brought up earlier that I wasn’t sure if I should be suspecting my CTF estimations for why some things aren’t turning out well. I think the answer is going to end up being that yes there are definitely issues with the CTF in a lot of the images, but a lot of the problems I’ve been asking about are probably a heterogeneity issue.