OmicVerse Single-Cell SCENIC

Goal

Run a reusable SCENIC analysis spine on single-cell AnnData: verify motif and ranking resources, initialize ov.single.SCENIC, infer a GRN, build regulons and AUCell scores, then optionally continue into RSS scoring, regulon binarization, and TF-centered GRN exploration. Keep this skill centered on one SCENIC result object rather than on the tutorial dataset.

Quick Workflow

1. Inspect the input AnnData, especially the raw-count layer name, gene symbols, and grouping column for downstream interpretation.
2. Confirm that the ranking databases and motif annotation table exist before constructing ov.single.SCENIC(...).
3. Pick the GRN branch up front with method='regdiffusion', method='grnboost2', or method='genie3'.
4. Run SCENIC.cal_grn(...) on a raw-count layer, then run SCENIC.cal_regulons(...) with explicit module-building kwargs when you need the notebook-style pruning behavior.
5. Treat the returned regulon AnnData plus scenic_obj.auc_mtx, scenic_obj.regulons, and scenic_obj.modules as the shared handoff point for all downstream stages.
6. Only if the user asks for interpretation, continue into RSS, binarization, embedding overlays, regulon marker ranking, or TF-target GRN plotting.
7. Validate the expected layers, tables, and object attributes before writing outputs.

Interface Summary