Gene Regulatory Network Inference (pySCENIC)

Gene Regulatory Network Inference (pySCENIC)

Infer gene regulatory networks (GRNs) de novo from single-cell RNA-seq data using pySCENIC. This workflow discovers transcription factor (TF) regulons directly from expression patterns and calculates cell-level TF activity scores.

When to Use This Skill

Use pySCENIC GRN inference when you need to:

• ✅ Discover TF-target relationships de novo from single-cell RNA-seq data
• ✅ Calculate cell-level TF activity scores for regulatory programs
• ✅ Identify cell-type-specific transcriptional programs
• ✅ Find key regulators driving cell state transitions or differentiation
• ✅ Compare regulatory programs across conditions, tissues, or species

Don't use this skill for:

• ❌ Bulk RNA-seq with few samples → Use functional-enrichment-from-degs or tf-activity workflows
• ❌ Quick TF activity from DE results → Use curated network approaches (faster, less computational)
• ❌ <500 cells → Insufficient for robust GRN inference
• ❌ Limited computational resources → Requires 16GB+ RAM, several hours runtime

Key Concept: Unlike curated network approaches, pySCENIC infers TF-target relationships directly from your data using co-expression analysis, then validates them using motif enrichment in cis-regulatory regions.

The SCENIC Pipeline:

1. GRN Inference (GRNBoost2): Identify co-expression modules linking TFs to potential target genes
2. Regulon Prediction (cisTarget): Prune targets to those with TF binding motifs in promoters
3. Cell Activity Scoring (AUCell): Score each cell for regulon activity

Quick Start

Fastest way to test the workflow (~10-15 minutes):

# Step 1: Load example PBMC data (500 cells)
from load_example_data import load_pbmc3k_example
adata, ex_matrix = load_pbmc3k_example(preprocess=True, subsample=500)

# Step 2: Run complete GRN workflow
from run_grn_workflow import run_complete_grn_workflow
results = run_complete_grn_workflow(
    ex_matrix=ex_matrix,
    tf_list_file="pyscenic_databases/allTFs_hg38.txt",
    database_glob="pyscenic_databases/*.feather",
    motif_annotations_file="pyscenic_databases/motifs-v10nr_clust-nr.hgnc-m0.001-o0.0.tbl",
    output_dir="scenic_test_results",
    n_workers=4
)

# Step 3: Visualize and export
from integrate_with_adata import integrate_with_adata
from plot_regulon_visualizations import generate_all_visualizations
from export_all import export_all

adata = integrate_with_adata(adata, results['auc_matrix'], results['regulons'])
generate_all_visualizations(results['auc_matrix'], results['regulons'], adata,
                            output_dir="scenic_test_results/plots")
export_all(results['regulons'], results['auc_matrix'], results['auc_summary'],
           results['adjacencies'], output_dir="scenic_test_results")