omicverse-single-cell-scenic

What problem does it solve?

This Skill automates reproducible SCENIC regulon inference and downstream regulon-level interpretation on single-cell AnnData objects, removing manual branch selection, resource checks, and brittle notebook steps.

Core Features & Use Cases

• GRN inference with branch selection: choose regdiffusion, grnboost2, or genie3 and run SCENIC.cal_grn with validated layer handling.
• Regulon construction and scoring: build modules_from_adjacencies, prune with cisTarget resources, compute AUCell matrices, and return a regulon AnnData handoff.
• Downstream interpretation: compute regulon specificity scores, binarize activities, rank regulon markers, and optionally visualize TF-target networks for targeted exploration.
• Use case: convert an OmicVerse SCENIC tutorial into a reusable pipeline that accepts an AnnData with raw counts and external cisTarget databases to produce AUCell, regulon binary calls, and RSS tables for downstream analysis.

Quick Start

Run SCENIC on my AnnData using regdiffusion with the counts layer and my local cisTarget databases to produce regulons, AUCell scores, and binarized activity calls.