Sliding window

[FrancescaDr]

Description

This PR adds a new function, sliding_window, to the codebase. The sliding_window function divides a tissue slice into regularly shaped, spatially contiguous regions (windows). It allows for the analysis of spatial data by segmenting the tissue into smaller, manageable windows with optional overlap.

How has this been tested?

PyTest on sq.datasets.mibitof()

Closes

Closes #829

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 76.13636% with 21 lines in your changes missing coverage. Please review.

Project coverage is 68.79%. Comparing base (4a632d6) to head (b16e8d5). Report is 47 commits behind head on main.

Files with missing lines	Patch %	Lines
src/squidpy/tl/_sliding_window.py	76.13%	11 Missing and 10 partials :warning:

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #859      +/-   ##
==========================================
- Coverage   69.99%   68.79%   -1.20%     
==========================================
  Files          39       39              
  Lines        5532     5660     +128     
  Branches     1037      907     -130     
==========================================
+ Hits         3872     3894      +22     
- Misses       1367     1459      +92     
- Partials      293      307      +14     

Files with missing lines	Coverage Δ
src/squidpy/tl/_sliding_window.py	76.13% <76.13%> (ø)

... and 10 files with indirect coverage changes

---- 🚨 Try these New Features:

• Flaky Tests Detection - Detect and resolve failed and flaky tests

[LLehner]

Hi @FrancescaDr, i just tested your code on data with a single slide (so I didn't provide a library_key) and it seems like the sliding windows are only created when you provide a library key.

E.g. the following code does run but doesn't return multiple sliding windows:

adata = sq.datasets.seqfish() sq.tl.sliding_window(adata, window_size = x) # here x can be any integer, doesn't change the result

[FrancescaDr]

Hi @FrancescaDr, i just tested your code on data with a single slide (so I didn't provide a library_key) and it seems like the sliding windows are only created when you provide a library key.

E.g. the following code does run but doesn't return multiple sliding windows:

adata = sq.datasets.seqfish() sq.tl.sliding_window(adata, window_size = x) # here x can be any integer, doesn't change the result

If I choose x = 1,2,3 I get a adata.obs['sliding_window'] with 'None_Z' where Z is the number of the sliding window.

[LLehner]

If I choose x = 1,2,3 I get a adata.obs['sliding_window'] with 'None_Z' where Z is the number of the sliding window.

Ah right, for this dataset the coordinates are on a small scale, i initially tried it with window size 10, 100, 1000...ok i see that it works now!

[FrancescaDr]

Ah right, for this dataset the coordinates are on a small scale, i initially tried it with window size 10, 100, 1000...ok i see that it works now!

Yes the scale sliding window size variable always depends on the density and size of the FOV. It might be valuable to add a function that suggest appropriate sliding window sizes depending on e.i. number of average cells in a window.