Chaotic Spatial

[DruncBread]

Hi,

I tried moscot on two DLPFC slices 151673 and 151507 ( 151673 as reference).
But after running the codes below for alignment, I found that the resulted spatial is messed up.
Would you mind helping me solve this problem?
Thank you!

adata = ad.concat([target,source])

ap = AlignmentProblem(adata=adata)
ap = ap.prepare(batch_key="batch", policy="sequential")
ap = ap.solve()

mode='affine'
ap.align(reference='151673' , mode=mode, spatial_key="spatial", key_added=f"spatial_{mode}")
sq.pl.spatial_scatter(
    adata, shape=None, spatial_key=f"spatial_{mode}", library_id="batch", color="batch"
)

I visualized the slices and found that the domain is messed up.

[MUCDK]

Hi,

Could you please clarify what you mean by "messed up"? What do you expect? How do you conclude that it's not what you want it to do?

Thanks!

[DruncBread]

Hi,

Sorry about the unclear description!

So I am attempting to align two DLPFC slices (151507 to 151673) using affine transformation. ( 151673 as reference, 151507 aligned to 151673 ).

However, after running the code above, the result was not as expected

As you can see in the code, I used affine transformation. So the aligned slices should have been simply rotated, but instead, the spots' domain (obs['ground_truth']) is incorrect, and it seems that the spots have been mixed up in the transformation.

I failed to upload the DLPFC dataset files on github. Would you mind leave your email so I could send you the data?

Thank you!

Best regards.