Tab Co occurrence

Co-occur

The Co-occur tab analyses how frequently pairs of cell clusters appear together across a range of spatial distances, producing a co-occurrence score as a function of distance that reveals at what scales clusters are associated or mutually exclusive.

Controls

Control	Description
Clustering	Clustering to analyse
Distance bins	Slider (10–100, default 50) — number of distance bins to compute
Run Co-occurrence	Computes the co-occurrence array across all distance bins
Cluster checkboxes	Three-column grid — select which source clusters to include in the plot
Select All	Checks all cluster checkboxes
Deselect All	Unchecks all cluster checkboxes
Filter targets	When checked, uses the active cluster selection from Cell Coloring to filter which target clusters appear in the plot
Show Co-occurrence Plot	Displays subplots for each selected source cluster showing co-occurrence vs. distance for all target clusters; enabled after running. Also saves the figure — see the notes.
Export CSV...	Saves a table with columns source_cluster, target_cluster, distance, co_occurrence; enabled after running

Workflow

1. Select a Clustering from the dropdown.
2. Set the number of Distance bins (more bins give finer resolution but take longer).
3. Click Run Co-occurrence and wait for completion.
4. Use the cluster checkboxes to choose which source clusters to visualise, then click Show Co-occurrence Plot.
5. Optionally enable Filter targets to restrict target clusters to those currently selected in Cell Coloring.
6. Click Export CSV... to save the co-occurrence table.

Notes

• A co-occurrence score greater than 1 indicates the cluster pair appears together more than expected by chance at that distance; a score less than 1 indicates avoidance.
• Results are persisted to adata.uns['co_occurrence'] and restored automatically on the next load.
• Show Co-occurrence Plot also writes the figure to <dataset>/plots/co_occurrence.<fmt>, where the format follows Preferences → Plot format (SVG by default, PNG at 300 dpi if selected). No save dialog appears; the file is simply written.
• Export CSV... writes only the source clusters currently ticked, not every cluster in the clustering.
• A progress bar under Run Co-occurrence tracks the computation, which is the slow part on a large section.