← Về thư mục
📄 / / profiles / r-and-d / raw_evidence / OpenAI4S / skills / bioskills / bio-gene-regulatory-networks-multiomics-grn / references / usage-guide.md

Multiomics GRN Inference - Usage Guide

Overview

Build enhancer-driven gene regulatory networks (eGRNs) by integrating single-cell RNA-seq and ATAC-seq. SCENIC+ discovers eRegulons: triplets linking transcription factors to target enhancers and downstream genes. The core advance is that scATAC nominates the actual distal enhancers active in the cells, so an edge is a TF -> region -> gene triplet with the region as a validatable anchor. But every inference arrow leaks (motif != binding, accessible != this-TF, peak-gene correlation != control), the peak-to-gene linking step is confounded by cell-type composition, and metacell aggregation introduces pseudo-replication -- so an eGRN is a prioritized hypothesis list, not a wiring diagram. Method choice hinges on paired vs unpaired data and on whether the downstream goal is perturbation simulation (build a CellOracle base GRN here, simulate in perturbation-simulation).

Prerequisites

# SCENIC+ and dependencies
pip install scenicplus pycisTopic

# Peak calling
pip install macs3

# Additional
pip install scanpy loompy matplotlib seaborn

# Download cisTarget databases for SCENIC+
# See: https://resources.aertslab.org/cistarget/

# FigR alternative (R)
devtools::install_github('buenrostrolab/FigR')
BiocManager::install(c('Signac', 'Seurat'))

Quick Start

Tell your AI agent what you want to do: - "Infer gene regulatory networks from my 10x multiome data" - "Run SCENIC+ on my paired scRNA+scATAC dataset" - "Find eRegulons linking TFs to enhancers and target genes" - "Build a multiomics GRN from my CellRanger ARC output"

Example Prompts

SCENIC+ Pipeline

"I have 10x multiome data processed with CellRanger ARC. Run SCENIC+ to identify eRegulons."

"Infer enhancer-driven regulatory networks from my paired scRNA-seq and scATAC-seq data."

Interpretation

"Which TFs have the most enhancer targets in my eRegulon results?"

"Show activating vs repressive eRegulons for each cell type."

FigR Alternative

"Run FigR on my Seurat multiome object to find TF-gene regulatory links via DORCs."

Unpaired or RNA-only

"I have separate scRNA-seq and scATAC-seq experiments, not paired. Integrate them with GLUE before inferring the GRN."

"I only have scRNA-seq. Build a GRN using a prebuilt CellOracle base GRN."

What the Agent Will Do

  1. Load paired scRNA-seq and scATAC-seq data
  2. Call peaks from ATAC fragments with MACS3
  3. Run cisTopic for topic modeling on scATAC regions
  4. Link accessible regions to target genes (region-to-gene)
  5. Connect TFs to enhancers via motif enrichment (TF-to-region)
  6. Assemble eRegulons (TF-enhancer-gene triplets)
  7. Score eRegulon activity per cell and visualize

Tips