squidpy API equivalence#

The public API — squidpy.experimental.tl and the two experimental.im helpers it needs — is covered by these notebooks. Upstream’s own notebooks call STalign.LDDMM and friends directly and build their rasters, coordinate axes and initial affines by hand along the way; these are sixteen of those seventeen analyses written against the public surface instead, one page per upstream notebook and named after it, so ours can be read beside STalign’s own.

Nothing here reaches into a private module. Four calls cover every case:

call

when

align_landmarks

paired points, closed-form affine, no iteration

stalign_align_obs

two point clouds; it rasterizes both sides itself

stalign_align_image

one side is an image, so the other is rasterized onto its grid

stalign_align_volume

a 2D section into a 3D reference volume

Each notebook links the upstream analysis it mirrors, and every function it calls links to its source. Where a notebook departs from upstream — a different raster scale, a different starting affine, curves resampled so they can be paired at all — it says so at that point in the notebook rather than in a list here.

A section into a reference volume#

Two point clouds#

Points onto an image#

The one upstream notebook without an equivalent here is merfish-merfish-alignment-simulation, which fits a section against a synthetically deformed copy of itself rather than against another section.