Hey everyone, i’m trying to analyse interaction of a protein with a lipid that is clearly seen in an important hotspot on a protein surface. I can clearly identify a lipid class, but lack important details. Are there any techniques specific to this case that would allow me to squeeze more information from it?
Membrane protein is fairly rigid and large, ie 1 conformation on a grid and >100 kDa.
I tried refining with masks for slightly-greater-than-protein region, but they don’t help much, especially given that’s its an oligomeric assembly.
Does 3DVA show the lipid moving at all (if it is included in the mask - if not, try 3DVA with a mask encompassing the protein + lipid)?
Have you tried 3D classification w/o alignments using a tight mask around the feature of interest? The parameters and mask size will likely need some optimization but it can help sort out heterogeneity at a region of interest.
- For mask generation, I start with a PDB and map in chimera, select the ligand of interest, then use the find clash tool (tools, structure analysis, find chlashes/contact) and set a radius of -5Å (optimize this) to select nearby atoms, then invert the selection, delete non-target regions, molmap the region of interest to 8Å, vop resample ongrid map, mask create in relion or cryosparc with generous soft edge (8Å).
On the post-processing side, you can try density modification in phenix for ligand/lipid enhancement.
If the lipid is lower resolution compared to the rest of the map, or heterogeneous, filtering by local resolution can help visualize it. The previously non-existent density becomes visible in the local filtered map at lower threshold and informs on whether focused 3D classification methods might be worthwhile.
Also, local refinement using a mask on the TM domain. If multiple TM domains are present, try focusing on the lipid-containing domain(s).
2 Likes
Does 3DVA show the lipid moving at all (if it is included in the mask - if not, try 3DVA with a mask encompassing the protein + lipid)?
I would say it doesn’t, but I’ll check additionally.
For mask generation, I start with a PDB and map in chimera, select the ligand of interest, then use the find clash tool (tools, structure analysis, find chlashes/contact) and set a radius of -5Å (optimize this) to select nearby atoms, then invert the selection, delete non-target regions, molmap the region of interest to 8Å, vop resample ongrid map, mask create in relion or cryosparc with generous soft edge (8Å).
Are there any particular considerations if a symmetry (C3) is present? Should I select only one subunit for a mask, or multiple?
On the post-processing side, you can try density modification in phenix for ligand/lipid enhancement.
Yeah, doesn’t help much.
I should have started by asking, regarding
“I can clearly identify a lipid class, but lack important details.”
What do you mean by lipid class? Does this mean some type of lipid species is clearly present in the consensus refinement, or there are multiple classes present and you can clearly separate particles for a single class with some kind of lipid density but the density still needs work?
What important details do you lack? How does the density look? Is it broken and discontinuous? Is it large and blobby compared to the rest of the map?
If you lowpass filter the map to 1Å worse than the reported resolution does the lipid become any more interpretable?
At what sigma level is the lipid clearly observable?
Are there any particular considerations if a symmetry (C3) is present? Should I select only one subunit for a mask, or multiple?
If your refinement is done with C3 symmetry imposed, you can put a mask on the lipid region in a single protomer and run symmetry expansion as input for 3D classification. Then your 3D classification will look at all that location in all protomers at the same time. If you find a good class, then homogeneous reconstruct C1 symmetry or local refine C1 symmetry - this will hopefully maintain asymmetry with biological relevance / class separation at the classification site in the context of the whole molecule. Perhaps other regions will have altered conformation.
Also try remove duplicates and homogeneous refine with C3 symmetry - be careful imposing symmetry but technically it’s ok if the results look ok as long as you removed duplicates to undo symmetry expansion.