Align 3D Maps, apply mask for alignment but not for output? (feature request)

Utilities
1

Hi,

I have a volume with pseudosymmetry, so when I use Align 3D Maps, it’s helpful to mask both the reference volume and the aligned volume on the matching “ASUs” so that the algorithm knows which subunits to pair with each other. However, I don’t want the output map to be masked. Currently, the options to use the mask for alignment and for the output volume are bundled together in a single switch. Instead, I’d like to have two separate buttons, so that I can say “yes, use the mask for computing the best alignment” and “no, do not multiply the transformed volume by the mask.”

Thanks!
-Josh

1 Like
2

Can you try to do a “Homogenous Reconstruction Only” job after aligning the map? Then it should be un-masked again, no?

1 Like
3

That’s a good idea that I tried already, but it doesn’t actually produce a transformation of the original volume. I can see this by eye, and here’s my best guess of what’s causing the difference. My original volume was generated by non-uniform refinement with marginalization turned on. So,

  1. @mmclean confirmed before that Homogeneous Reconstruction Only does not marginalize over poses (Homogeneous reconstruction only (2) - #4 by mmclean)
  2. Here’s another relevant question I’ve always wondered about: in non-uniform refinement jobs, is the space-varying regularizer applied to the output volume at the very end, or is it only applied to the reference volume during pose marginalization?

For now, I had GPT write me a script that will convert the rotation matrix / translation vector output by Align 3D Maps into “re-centering” coordinates and euler angles that I can then feed into Volume Alignment Tools.

4

Hi Josh - are you sure you have update particle orientations on? If you do then homogeneous reconstruction ought to give the shifted volume (if it didn’t align 3d –> local refinement would not work, and it does) - you definitely shouldn’t need to do any manual conversion of the rotation matrix for this

5

Hi Oli,

I should clarify what I meant by “it doesn’t actually produce a transformation of the original volume.” Homogeneous Reconstruction Only does successfully produce a very similar volume in the intended new location, but it is not actually a transformed copy of the original volume generated by the non-uniform refinement. Instead it is a transformed copy of the volume that would have resulted if I’d run Homogeneous Reconstruction Only on the original output of my non-uniform refinement job—and I think this is due to the mentioned difference(s?) between the way non-uniform refinement and homogeneous reconstruction generate maps. The very slight differences would not matter if the particle poses were what I was after (as in the Align 3D → local refinement pipeline that you mentioned), but in this case I’m after the volume itself (for things like model building and deposition). Does that make sense?

By the way, what I’m actually working on is taking the result of a local refinement (with non-uniform refinement on) on a subvolume that I previously moved to the center of the box and moving it back out to the position where it sat in the original consensus reconstruction. The reason I put it in the center of the box in the first place was your comment from several years ago here (Manual definition of refinement alignment center - #6 by olibclarke). Is it still your preferred workflow to move messy peripheral domains to the center of the box for local refinement?

6

oh right, I see what you mean… yes that makes sense!

For centering domains for refinement yes we use volume alignment tools - usually just by creating a mask around the region of interest and using VAT to center that

1 Like