Hi all,
What is the best way to analyze and compare two similar maps (protein+ligand vs apo-protein)? The main idea is to demonstrate that the additional density observed is indeed the ligand and also to visualize the subtle conformational changes between the two maps.
I processed the datasets independently (NU refinement - local refinement - DeepEM enhancer) and then used Align3D to align the sharpened maps and get the difference map. The maps are at similar resolution: 3.4 Å and 3.6 Å; the same box size and pixel size (collected on the same microscope). Is this the correct strategy or is there a better way to do it? Thank you!
If the conformational change is minimal, it might be worthwhile running a joint refinement of both sets of particles, and then just reconstructing both sets separately (using the consensus poses). This has a couple of advantages - firstly, it effectively doubles your number of particles, and therefore should improve the overall resolution of the consensus somewhat, and secondly, your two reconstructions will be well aligned to begin with.
If the ligand binding is associated with some degree of conformational change near the ligand binding site, you might want to calculate a locally aligned difference map (to show just difference density for the ligand absent domain reorientation). I do this using an alias in UCSF Chimera which you can find here.
It looks like there is a change in the relative orientation of the domains in the protein-ligand structure, plus some changes in the loops near the ligand. I tried running a joint refinement of both sets:
NU refinement;
Local refinement;
Curate exposures to separate the aligned particles back to the respective datasets;
Homogeneous reconstruction only
but I did not see the expected improvement; the resolution of joint refinement was the same (3.4 Å), and the maps appear noisier (both the consensus and two reconstructions). I assume because of that conformational change the particles could not align well?
I will certainly try a locally aligned difference map, thank you for the tip!
but I did not see the expected improvement; the resolution of joint refinement was the same (3.4 Å), and the maps appear noisier (both the consensus and two reconstructions). I assume because of that conformational change the particles could not align well?
Yes this sounds reasonable to me - if there is a substantial conformational change it won’t work well
Perhaps breaking it down into domains and carrying out local refinements would work better than the global refinement?
Had a domain which was stuck at ~4.5-5A (bumpy sausage alpha helices) get all the way to 2.6 (some really nice side chains) last week this way. Then 3D class to tease out two slightly different conformations.
In addition to these suggestions, I would recommend to check what you get without using deepEMhancer, because it tends to attenuate non-protein features (due to its training set containing only protein models). So you might have better ligand density in the experimental map than in the deepEMhancer map.