Without knowing more about your complex, this will be some sort of shot in the dark.
Well, dynamics is always present some way or another, and for different reasons. But to me it could also be that part of your particles do not have the compound bound. You might try to mask the compound and sourrounding residues - use a rather large region - and run masked 3D classification to try to separate bound from unbound. In my hands this work best when I also provide a static solvent mask. Depending on several factors - including the number of particles and signal-to-noise ratio of the images - this might work or not. Try different high resolution cuts like 12, 8, 6, 4 angstroms.
Another possibility that I’d like more experienced people to comment on, is if small compounds are always more susceptible to damage by the beam. In that case it should happen to pretty much everyone doing drug design with CryoEM…
Yes, I noticed that a lot of cryo-EM maps, especially for small-molecules drug designing, had the same issue. It is a good idea to try another run of local-3D classification in high resolution. Of course, it could also be the possibility of the beam damage…
Doesn’t seem to be the case, but I’m sure there are related publications. well-ordered small molecules have excellent density, often better than local region of the protein. Could test for yourself, though laborious, could use only early frames or only late frames of this exact particle set and see if they’re different, but both would yield an even worse structure (since you’re using less data). Your advice above is good, see takanori first particle subtraction relion publication as example of the same strategy.
Likely flexible ligand. Your map doesn’t seem quality enough to “confirm the ligand”. higher resolution structural information will be very helpful. consider selecting better CTF, running NU-refine with higher order aberration correction, run Ref Based MC, try a dozen local-refinement jobs with masks of varying sizes (and minimal deviations in shifts/rots and recenter shifts/rots each iteration) etc.
