single-scenic-grn_skill

Overview

This skill implements SCENIC gene regulatory network analysis inside OmicVerse for bulk, single-cell, and spatial RNA-seq. It infers TF→target links with RegDiffusion, prunes regulons by cisTarget motif enrichment, and scores regulon activity per cell with AUCell. The output includes regulons, AUCell activity matrices, RSS for cell-type specificity, and visualizations for embedding and GRN graphs.

How this skill works

The pipeline runs three stages: (1) GRN inference using RegDiffusion on raw count data to predict TF→target importance, (2) regulon pruning with cisTarget motif rankings and motif-to-TF annotations to keep direct targets, and (3) AUCell scoring to quantify regulon activity in each cell. The skill validates database presence, checks raw-count layers and gene-name conventions, and stores adjacencies, regulons, AUCell matrices, and intermediate modules for downstream analysis.

When to use it

• Inferring transcription factor regulatory networks from single-cell or bulk RNA-seq counts.
• Identifying regulons (TF + direct target gene sets) and master regulators for cell types.
• Scoring regulon activity per cell for regulon-based clustering or cell-type annotation.
• Comparing regulon specificity across annotated cell types using RSS.
• Visualizing TF activity on embeddings and plotting TF→target networks for interpretation.

Best practices

• Provide raw (non-log) counts in a named layer (e.g., adata.layers['counts']) — RegDiffusion requires count-level variance.
• Download species-appropriate cisTarget .feather ranking databases and motif .tbl files before running; missing DBs are the most common failure.
• Filter HVGs to ~2000–3000 genes for tractable runtime and memory; reduce HVGs if pruning causes MemoryError.
• Ensure gene name case matches the database species (mouse mixed-case, human uppercase) to avoid empty regulons.
• Set n_jobs to match available CPU cores; reduce parallelism if memory is constrained.
• Validate existence of db_glob with glob.glob and check motif_path file before starting.

Example use cases

• Run SCENIC on mouse hematopoiesis scRNA-seq to find master regulators per annotated cell type.
• Infer GRNs from spatial transcriptomics counts and map regulon activity onto tissue embeddings.
• Score and binarize regulon activity to identify on/off programs for downstream differential activity tests.
• Prune co-expression modules to direct regulons and visualize top TF→target edges for publication figures.
• Compute RSS to rank TFs specific to rare cell populations for follow-up experiments.