OmicVerse Single-Cell Differential Abundance

Goal

Turn notebook-style OmicVerse differential cell-type abundance analysis into a reusable execution spine. Keep this skill separate from DEG because DCT has a different constructor, different branch-specific prerequisites, and different result shapes.

Quick Workflow

1. Inspect the input AnnData, the condition column, the cell-type column, and whether sample IDs and a usable embedding already exist.
2. Choose a backend before running anything: sccoda for the scCODA branch, milopy for the native Python Milo-like branch, or milo for the pertpy Milo branch.
3. For milopy and milo, make sure sample_key exists and use_rep points to an embedding already stored in obsm.
4. Construct ov.single.DCT(...) with explicit branch arguments, then run run(...).
5. For sccoda, do any posterior-specific follow-up such as set_fdr(...) only after the model has finished.
6. Collect results with get_results(...) and verify the branch-specific output columns before plotting or exporting.

Interface Summary