Co-accessibility (cis-Regulatory Linkage) - Usage Guide

Overview

Infer enhancer-promoter and enhancer-enhancer cis-regulatory connections from scATAC-seq using cell-to-cell variability in joint peak accessibility. Cicero, ArchR getCoAccessibility, and SCENIC+ produce peak-pair connection scores; thresholding generates enhancer-gene candidates. Co-accessibility is a hypothesis generator -- it correlates with Hi-C/Micro-C 3D contacts at ~30-50% but is not equivalent. Validate against orthogonal data (Hi-C, CRISPRi-FlowFISH).

Prerequisites

BiocManager::install(c('GenomicInteractions', 'BSgenome.Hsapiens.UCSC.hg38'))
# monocle3 and the monocle3-compatible Cicero are GitHub-only (the Bioconductor `cicero` targets monocle2)
remotes::install_github('cole-trapnell-lab/monocle3')
remotes::install_github('cole-trapnell-lab/cicero-release', ref='monocle3')
remotes::install_github('GreenleafLab/ArchR', ref='master')

# SCENIC+ for TF-driven networks (Multiome only); source-only install, pulls in pycisTopic
git clone https://github.com/aertslab/scenicplus && cd scenicplus && pip install .

Inputs: scATAC peak-cell matrix (from Signac, ArchR, or SnapATAC2 preprocessing) and a UMAP/dim-reduction for metacell construction. SCENIC+ additionally requires paired RNA AnnData (Multiome).

Quick Start

Tell your AI agent what you want to do: - "Run Cicero on a Signac scATAC-seq object to get peak-pair co-accessibility" - "Use ArchR getCoAccessibility for an ArchR project; connection cutoff 0.5" - "Find enhancer-gene candidates by overlapping connections with promoters in tssRegion=c(-2000, 500)" - "Compare Cicero connections against published Hi-C loops to estimate concordance" - "Run SCENIC+ on Multiome data to infer TF -> enhancer -> gene networks"

Example Prompts

Standard Cicero

"Convert my Signac scATAC peak matrix to a CDS, reduce dimensions with LSI + UMAP, build a Cicero CDS via metacells (k=50), then run_cicero with window=500000 for cis connections. Filter to coaccess > 0.25 as high-confidence."

Per-Cluster Co-accessibility

"Run Cicero separately on each Seurat cluster to capture cell-type-specific connections; merge results with cluster annotation."

ArchR Wrapped Cicero

"In my ArchR project, addCoAccessibility on PeakMatrix with k=100, knnIteration=500, maxDist=250000 (250 kb cis). Use corCutOff=0.5 for high-confidence."

Enhancer-Gene Linking

"Use the strong co-accessibility connections; map one anchor to gene promoters (TSS +/- 2 kb); the other anchor is the candidate enhancer. Output enhancer-gene pairs."

Hi-C Concordance

"Take all coaccess > 0.5 connections; overlap them with HiCCUPS loop calls from a published Hi-C dataset matching the cell type. Report the % of Cicero connections supported by Hi-C."

SCENIC+ TF Networks

"Run SCENIC+ on Multiome data: pycisTopic for ATAC topics, motif enrichment per topic (cisTarget), then SCENIC+ to integrate with paired RNA. Output TF-enhancer-gene triplets."

What the Agent Will Do

Verify input is scATAC peak matrix (from Signac/ArchR/SnapATAC2)
Reduce dimensions (LSI + UMAP) -- typically already done
Build metacell-aggregated CDS (Cicero) or use ArchR's k-NN aggregation
Run co-accessibility within window=500000 (cis) by default
Apply graphical lasso with default or estimated alpha
Filter to high-confidence connections (coaccess > 0.25 standard; > 0.5 stringent)
Optionally per-cluster runs to capture cell-type specificity
Map connections to genes via promoter overlap (TSS +/- 2 kb)
Compare against Hi-C / Micro-C if reference available
Output BEDPE for visualization

Tool Decision Quick Reference

Setting	Tool
ATAC-only standard	Cicero
ATAC-only inside ArchR	ArchR getCoAccessibility
Multiome (RNA + ATAC) direct linking	LinkPeaks (Signac)
Multiome TF networks	SCENIC+
Pre-computed reference	GeneHancer / FANTOM5 / EpiMap

Connection Score Quick Reference

Score	Use
> 0.5	Stringent / high-confidence
> 0.25	Standard reporting (Cicero vignette convention)
> 0.05	Exploratory; many false positives
< 0.05	Not biologically meaningful

Tips

Co-accessibility is NOT 3D contact. It's a statistical association from cell-to-cell co-variation. Hi-C concordance is typically 30-50%.
Run Cicero per-cluster for heterogeneous datasets; pooled metacells dilute cell-type-specific connections.
Default window=500000 (500 kb cis) excludes most distal regulation. Widen for gene desertless TADs; trans connections require Hi-C, not co-accessibility.
Cicero's alpha parameter dramatically shifts results; use estimate_distance_parameter() for data-driven choice.
Multiome data enables direct enhancer-gene correlation (LinkPeaks) which is more direct than co-accessibility alone.
SCENIC+ is multi-step and complex; budget 1-2 days for setup. Outputs are TF -> enhancer -> gene triples.
Co-accessibility predictions are hypotheses. Validate with Hi-C, ChIP-seq, or CRISPRi-FlowFISH.
Heterochromatic 3D contacts are constitutive (don't vary in accessibility) and won't appear in co-accessibility. This is expected, not a bug.
For published reference enhancer-gene pairs: GeneHancer (cell-type-agnostic), FANTOM5 (CAGE-based), EpiMap (~833 epigenomes).
The same biology gives different connection counts at different metacell k values; lower k = more variability captured but slower runtime.

atac-seq/single-cell-atac - scATAC preprocessing (input)
atac-seq/consensus-peakset - Peak set for connection inference
atac-seq/motif-deviation - chromVAR (complement)
gene-regulatory-networks/scenic-regulons - Standalone SCENIC for TF networks
hi-c-analysis/loop-calling - Hi-C physical contacts (validation)
hi-c-analysis/contact-pairs - Hi-C / Micro-C contact pairs
single-cell/multimodal-integration - Multiome integration
chip-seq/peak-annotation - Cross-validate with ChIP
pathway-analysis/gsea - Downstream gene-level enrichment