Skip to contents

Count strand-oriented insertion endpoints around ATAC regions

Source: R/ATACRegions.R
RunATACRegionMatrix.Rd

Build group-specific region-by-position matrices from local indexed fragments, without subsampling cells or regions. Repeated endpoints are counted with their multiplicity. This is an explicit SCOP endpoint counter, not a call to Signac's RegionMatrix.

Usage

RunATACRegionMatrix(
  object,
  regions,
  assay = NULL,
  group.by = NULL,
  cells = NULL,
  upstream = 3000L,
  downstream = 3000L,
  bin.size = 50L,
  max.entries = 5e+07,
  verbose = TRUE
)

Arguments

object

A Seurat object containing a validated ChromatinAssay with local indexed fragments and explicit cell-to-barcode mappings.

regions

Non-empty GRanges in 1-based closed coordinates. Each region is centered at start + floor((width - 1)/2). Minus-strand regions are reversed; unstranded regions use genomic orientation. Duplicate genomic intervals are rejected even when their names differ. Unique names are used as row IDs when provided, otherwise genomic coordinate IDs are used.

assay

Chromatin assay; defaults to the active assay.

group.by

Metadata column for grouping. NULL uses active identities.

cells

Unique source-assay cells; NULL includes all assay cells.

upstream, downstream

Non-negative integer distances from the center. Windows must fit entirely within known chromosome boundaries; no trimming or reference/chromosome conversion is performed.

bin.size

Positive integer width of positional bins. The last bin may be shorter; actual widths and centers are returned.

max.entries

Maximum number of dense output entries across all groups. Exceeding this explicit memory guard errors rather than sampling. This guards output size, not transient memory for a single dense fragment query.

verbose

Whether to print progress messages.

Value

A scop_atac_regions list with matrices (one raw count matrix per group), regions, bin positions/widths, cells.per.group, and a receipt.

Details

A BED fragment [start,end) contributes two endpoints, start + 1 and end, in 1-based coordinates. Only endpoints inside each region window are counted. Read-support column 5 is not a weight. There is no extra Tn5 shift, bias correction, fragment-length weighting or depth normalization. Overlapping windows can count the same endpoint. Grouping is descriptive, not a replicate-level statistical test; library-size and cell-composition confounding remain. Declared reference contigs absent from a file produce zero counts for that file, not evidence of biological inaccessibility. The caller's object is not modified.

See also

ATACRegionHeatmap(), ATACRegionPlot()