bio-atac-seq-co-accessibility

Version Compatibility

Reference examples tested with: Cicero 1.20+, monocle3 1.3+, ArchR 1.0.2+, SCENIC+ 1.0+, pycisTopic 1.0+, Signac 1.13+, GenomicRanges 1.54+, GenomicInteractions 1.36+, BSgenome.Hsapiens.UCSC.hg38 1.4+.

Verify before use:

• R: packageVersion('<pkg>') then ?function_name to verify parameters
• Python: pip show <package> then help(module.function) to check signatures

If code throws unexpected errors, introspect the installed package and adapt rather than retrying.

Co-accessibility (cis-Regulatory Linkage)

"Which enhancers connect to which promoters in my scATAC data?" -> Use cell-to-cell variability in joint accessibility of nearby peaks to infer cis-regulatory connections without explicit RNA expression. Output is a peak-pair graph with co-accessibility scores; thresholding produces enhancer-gene candidate pairs.

• R: cicero::run_cicero(input_cds, genomic_coords) -> peak-pair connection scores
• R: ArchR::addCoAccessibility(proj) -> ArchR-internal Cicero wrapper
• Python: pycisTopic + SCENIC+ for network-level inference combining ATAC + RNA + motifs

Co-accessibility is NOT 3D contact; it's a statistical association based on cell-to-cell co-variation. Strong co-accessibility correlates with Hi-C/Micro-C contacts (~30-50% concordance) but is not equivalent.

What Co-accessibility Captures vs What It Doesn't

For physical contact, use Hi-C, Micro-C, or PCHi-C. Co-accessibility is the chromatin-only proxy.

Algorithmic Taxonomy

Methodology evolves; verify against Pliner 2018 (Cicero), Bravo 2023 (SCENIC+), Nasser 2021 (ABC model alternative for enhancer-gene), and current Hi-C concordance benchmarks.

How Cicero Works (Conceptually)

Cell-to-cell variability is too sparse for direct correlation. Cicero solves this via metacells:

1. Reduce dimensionality (UMAP from input).
2. Build k-NN graph of cells.
3. Aggregate k cells into metacells (default k = 50).
4. Compute correlation in accessibility across metacells, restricted to peak pairs within genomic_distance_max (default 500 kb cis).
5. Apply graphical lasso with regularization alpha to sparsify the correlation matrix.
6. Output: per-pair connection score; positive = co-variation, negative = anti-co-variation.

Connection thresholds typically 0.05-0.5; > 0.25 is high-confidence.

Per-Tool Failure Modes

Cicero -- alpha tuning shifts results

Trigger: Default alpha (sometimes computed automatically from data); custom alpha < 0.5 or > 5.

Mechanism: Alpha controls graphical lasso regularization. Too low: dense graph with many spurious connections; too high: sparse with biology missing.

Symptom: Connection count varies 10-100x across alpha sweeps.

Fix: Use Cicero's estimate_distance_parameter() to get data-driven alpha; verify connection count is biologically plausible (~10-50% of peaks have at least one strong connection).

Cicero -- metacell aggregation hides cell-type-specific connections

Trigger: Running Cicero on heterogeneous dataset spanning multiple cell types.

Mechanism: Metacells aggregate across cell types; connections that exist only in one cell type get diluted.

Fix: Run Cicero per-cluster separately; combine results with cluster annotations. Cell-type-specific connections often differ.

Cicero -- distance assumption

Trigger: Default genomic_distance_max=500000 (500 kb cis only).

Mechanism: Distal connections beyond 500 kb cis are excluded; trans-chromosomal entirely missed.

Fix: For specific use cases (e.g., gene desertless TADs), increase genomic_distance_max to 1 Mb or more. Trans connections require Hi-C, not co-accessibility.

SCENIC+ -- RNA scaling

Trigger: RNA-side dropouts in Multiome data.

Mechanism: SCENIC+ requires reasonable RNA quantification per cell. Sparse Multiome RNA with many zero genes causes correlation degradation.

Fix: Filter cells with insufficient RNA; aggregate cells if necessary. Multiome RNA should look comparable to standalone scRNA-seq.

LinkPeaks (Signac) -- Distance default

Trigger: Default LinkPeaks(..., distance=5e+05).

Mechanism: Same as Cicero; 500 kb cis only by default.

Fix: Same; widen if needed but trans not supported.

Decision Tree by Goal

Cicero Standard Workflow

Goal: Infer cis-regulatory peak-peak connections from a scATAC peak-cell matrix.

Approach: Build a Monocle3 CellDataSet, reduce dimensions via LSI + UMAP, aggregate cells into metacells, then run Cicero's graphical-lasso correlation across the cis window and threshold on connection score.

library(cicero); library(monocle3); library(GenomicRanges)

# Input: peak-cell binary matrix from Signac/ArchR (rows = peaks, cols = cells)
# Convert peaks to "chrN_start_end" format
peak_names <- paste0(seqnames(peaks), '_', start(peaks), '_', end(peaks))
input_cds <- new_cell_data_set(peak_matrix, cell_metadata=metadata,
                               gene_metadata=peak_metadata)

# Reduce dimensionality (UMAP from input)
input_cds <- detect_genes(input_cds)
input_cds <- estimate_size_factors(input_cds)
input_cds <- preprocess_cds(input_cds, method='LSI')
input_cds <- reduce_dimension(input_cds, reduction_method='UMAP',
                              preprocess_method='LSI')

# Build metacell-aggregated CDS
umap_coords <- reducedDims(input_cds)$UMAP
cicero_cds <- make_cicero_cds(input_cds, reduced_coordinates=umap_coords, k=50)

# Run Cicero with hg38 chrom sizes
genome_df <- data.frame(chr=seqnames(seqinfo(BSgenome.Hsapiens.UCSC.hg38)),
                        length=seqlengths(seqinfo(BSgenome.Hsapiens.UCSC.hg38)))
conns <- run_cicero(cicero_cds, genomic_coords=genome_df,
                    window=500000, sample_num=100)

# Filter to high-confidence connections.
# Threshold 0.25 is a Cicero-documentation working default; the optimal cutoff
# is dataset-dependent and is best calibrated against orthogonal Hi-C / HiChIP.
strong <- conns[conns$coaccess > 0.25, ]
cat(sprintf('Total conns: %d; strong (>0.25): %d\n', nrow(conns), nrow(strong)))

ArchR getCoAccessibility

library(ArchR)
proj <- loadArchRProject('ArchR_out')
proj <- addCoAccessibility(proj, reducedDims='IterativeLSI',
                          k=100, knnIteration=500,
                          maxDist=250000)               # 250 kb cis (tighter than default)
co_acc <- getCoAccessibility(proj, corCutOff=0.5,       # Default 0.5 in ArchR; lower for more (calibrate against Hi-C/HiChIP)
                             returnLoops=FALSE)

ArchR returns connections as a GRanges object compatible with GenomicInteractions for direct overlap with Hi-C loops.

Visualizing Connections

# As arc plot at a locus of interest
library(Gviz); library(GenomicInteractions)
gi <- GenomicInteractions(anchorOne=GRanges(strong[, 1:3]),
                          anchorTwo=GRanges(strong[, 4:6]),
                          counts=as.integer(strong$coaccess * 100))
plotInteractions(gi, view='arc', interactions.color='red')

For genome-browser visualization with ArchR: plotPeak2GeneHeatmap() shows the peak-gene linkage matrix; plotBrowserTrack() overlays connections on tracks.

SCENIC+ TF-Driven Networks

import scenicplus as sp

# Multi-step: requires preprocessed pycisTopic models + paired RNA AnnData
scplus_obj = sp.create_SCENICPLUS_object(
    GEX_anndata=rna_adata,
    cisTopic_obj=topic_obj,
    menr=motif_enr_dict)

# Calculate eRegulons (TF + target genes + linked enhancers)
sp.run_scenicplus(scplus_obj, ...)
scplus_obj.uns['eRegulon_metadata']

SCENIC+ is significantly more complex than Cicero; budget 1-2 days for setup. The benefit is that outputs are TF -> enhancer -> gene triples, not just peak-peak co-accessibility.

Cicero Alpha Mathematics

Trigger: Tuning Cicero's regularization parameter for the graphical lasso step.

Mechanism: estimate_distance_parameter() fits a regression of pairwise correlation versus genomic distance for nearby peaks; the slope gives the expected co-accessibility decay with distance. Cicero's alpha penalty is set proportional to this slope so the graph sparsifies at biologically appropriate distance scales.

Implementation: Cicero internally calls estimate_distance_parameter(cicero_cds, window=window, maxit=100, sample_num=100) -- the function bootstraps 100 windows, fits per-window distance-correlation regression, and averages the alpha estimates. Default value is robust at typical peak densities; manual override is rarely needed.

When manual tuning helps: Very dense peaksets (>200k peaks) may need higher alpha to control false positives; very sparse (<10k peaks) may need lower alpha to recover signal. Verify by running on technical replicates -- the expected outcome is ~0 strong connections.

ABC Model Cross-Reference

For enhancer-to-gene linking with paired Hi-C/Micro-C, the canonical method is the ABC model (Fulco 2019, Nasser 2021), not Cicero. ABC computes ABC = (Activity_E * Contact_E,G) / sum_e(Activity_e * Contact_e,G); standardizes on combined ATAC + H3K27ac activity and Hi-C contact frequencies. ENCODE-rE2G (2024) is the modern logistic-regression replacement.

See atac-seq/enhancer-gene-linking for full ABC and ENCODE-rE2G coverage. Cicero is the ATAC-only fallback when no Hi-C is available.

HiChIP H3K27ac as Orthogonal Anchor

Cicero is appropriate when no 3D data exists; do not use Cicero in lieu of ABC when Hi-C/Micro-C are available.

Hi-C / Micro-C Concordance

Goal: Quantify what fraction of strong Cicero connections are supported by Hi-C loop calls.

Approach: Import HiCCUPS loops as GenomicInteractions, build a parallel object from Cicero connections, then count anchor-anchor overlaps and report the percentage.

# Compare Cicero against published Hi-C loops
library(GenomicInteractions)
hic_loops <- import('hiccups_loops.bedpe', format='bedpe')
ci <- GenomicInteractions(anchorOne=GRanges(strong$Peak1),
                          anchorTwo=GRanges(strong$Peak2))
overlap <- countOverlaps(ci, hic_loops, type='equal') > 0
cat(sprintf('Cicero connections overlapping HiCCUPS loops: %.1f%%\n',
            100 * mean(overlap)))

Reconciliation

Operational rule: Co-accessibility is a hypothesis generator. Validate with Hi-C, ChIP-seq, or experimental enhancer-promoter interaction (CRISPRi-FlowFISH).

Common Errors

References

• Pliner HA et al 2018 Mol Cell 71:858 (Cicero)
• Granja JM et al 2021 Nat Genet 53:403 (ArchR getCoAccessibility)
• Bravo Gonzalez-Blas C et al 2023 Nat Methods 20:1355 (SCENIC+)
• Stuart T et al 2021 Nat Methods 18:1333 (Signac LinkPeaks)
• Nasser J et al 2021 Nature 593:238 (ABC model; alternative enhancer-gene)
• Fulco CP et al 2019 Nat Genet 51:1664 (CRISPRi-FlowFISH; gold-standard validation)
• Mumbach MR et al 2017 Nat Genet 49:1602 (HiChIP H3K27ac for enhancer-promoter)
• Boix CA et al 2021 Nature 590:300 (EpiMap; bulk enhancer-gene reference)

Related Skills

• atac-seq/single-cell-atac - scATAC preprocessing (input)
• atac-seq/consensus-peakset - Peak set used for connection inference
• atac-seq/motif-deviation - chromVAR for TF activity (complement)
• atac-seq/enhancer-gene-linking - ABC, ENCODE-rE2G, CRISPRi-FlowFISH validation when Hi-C is available
• atac-seq/deep-learning-atac - chromBPNet variant effect at predicted enhancers
• gene-regulatory-networks/scenic-regulons - Standalone SCENIC for TF networks
• hi-c-analysis/loop-calling - Physical contacts from Hi-C
• hi-c-analysis/contact-pairs - Hi-C / Micro-C contact pairs
• single-cell/multimodal-integration - Multiome integration
• chip-seq/peak-annotation - Cross-validate with TF ChIP
• pathway-analysis/gsea - Downstream gene-level enrichment