Adjacent Pairwise Alignment

[DruncBread]

Hi,

I wonder if there is any way I could perform adjacent pairwise alignment with moscot
Like slice1 aligned to slice0, slice2 aligned to slice1, slice3 aligned to slice2.....

It would be great if you could explain this to me! Thank you!

Best regards.

[MUCDK]

Hi,

This should work with the "sequential" policy.

Check out this example https://moscot.readthedocs.io/en/latest/notebooks/examples/problems/400_subset_policy.html and lmk if you have further questions.

[MUCDK]

@ArinaDanilina let's add a sequential and Starpolicy example to https://moscot.readthedocs.io/en/latest/notebooks/examples/problems/400_subset_policy.html for the AlignmentProblem as we currently only show examples in the temporal domain there

[DruncBread]

Thank you very much for your guidance!
I check out the "sequential" policy and it should work for me.
But I want to be sure if this means that slice1 will be aligned to slice0:
(0, 1): BirthDeathProblem[stage='prepared', shape=(20, 20)]

[DruncBread]

Hi,

later when I tried to transform the coordinate
ap.align(reference=sample_ids[0], mode="warp", spatial_key="spatial", key_added="spatial_warp_sequential")
Does that mean all the slices' coordinates will be aligned to slice0 (sample_ids[0]) ?

Is there any way to transform the coordinate sequentially?

[MUCDK]

Hi,

This is correct, with ap.align(reference=sample_ids[0], mode="warp", spatial_key="spatial", key_added="spatial_warp_sequential") you will align the coordinates of slice 1 to slice 0.

Can you please specify what you mean by transform the coordinate sequentially? I assume you have multiple slices (>2). In case you wanna map 1->2, 2->3, 3->4, etc. , you have two options:

• either instantiate three different AlignmentProblems, and instantiate those only with the relevant slides, and thus do pairwise alignments.
• or have only one AlignmentProblem with all slides and the sequential policy, then align aligns all other slides to the reference, e.g. 1->4, 2->4, 3->4, but 1->4 will not be aligned by using one transport map only (because you only computed transport maps sequentially, but it will be transformed by "chaining" mulitple transport maps).

Is this clear?
Pinging @giovp, do you have anything to add?

[DruncBread]

Thank you very much, MUCDK! It was really helpful!