Interval: windowed coverage around features

Use this recipe when you have point or summit features and want signal quantified in a fixed neighborhood around each — reads in summit ± N bp, coverage in promoter windows. The atacseq workflow uses it to count reads in 1 kb windows around MACS2 summits. Three operations, in order.

1. Window the features (slop)

bedtools_slopbed extends each summit by a fixed amount on both sides to make a window (“get summits +/-500kb”; addition: { addition_select: b, b: "500" } — 500 bp each side, a 1 kb window). Requires a genome file to clamp at chromosome ends. See interval-window-flank for parameters and the bp-vs-kb label-drift gotcha.

2. Collapse overlapping windows (merge)

Adjacent summits produce overlapping windows; bedtools_mergebed collapses them so a region of read pileup isn’t counted twice (“Merge summits +/-500kb”; distance: "0"). See interval-merge-overlapping.

3. Count reads in each window (coverage)

bedtools_coveragebed counts how many reads (from the deduplicated-reads BAM) fall in each merged window (“Compute coverage on summits +/-500kb”; reduce_or_iterate_selector: iterate, hist: false, mean: false). This is interval-coverage Mode B — region-restricted counts, not genome-wide depth.

Why this shape

The merge between slop and coverage is the non-obvious step. Without it, overlapping windows double-count shared reads and inflate the per-window signal. Slop defines the neighborhoods, merge makes them disjoint, coverage quantifies — each step is one operation, and dropping merge silently biases the numbers upward.

Pitfalls

• Skipping merge double-counts. This is the recipe’s whole reason to exist as a three-step chain rather than slop→coverage. Overlapping windows share reads; coverage over un-merged windows over-reports.
• Window size vs label. b: "500" is 500 bp; the step’s “500kb” label is wrong (see interval-window-flank). Verify the actual extension against the biology you want.
• Coverage mode. Use interval-coverage Mode B (coveragebed, reads-in-regions) here, not Mode A (genomecoveragebed, genome-wide) — you already have the regions and want per-region counts.
• Genome file consistency. Slop’s genome file must match the alignment reference, or windows land at coordinates the reads can’t occupy.

See also

• galaxy-interval-patterns — the interval pattern map.
• interval-window-flank — the slop step.
• interval-merge-overlapping — the merge step.
• interval-coverage — the coverage step (Mode B).
• iwc-interval-operations-survey — corpus evidence.