Subtract Bed

# 2) Remove whole A features if at least half the feature overlaps B
subtractBed \
  -a exons.bed \
  -b repeats.bed \
  -A \
  -f 0.5 \
  > exons.filtered.bed

# 3) Subtract only same-strand overlaps, treating BED12/BAM splits separately
subtractBed \
  -a transcripts.bed12 \
  -b antisense-mask.bed \
  -s \
  -split \
  > transcripts.trimmed.bed

Recommended Workflow

1. Decide whether you want partial trimming of A (default) or whole-record removal with -A / -N.
2. Add overlap thresholds and strand rules deliberately with -f, -F, -r, -e, -s, or -S.
3. For large datasets, coordinate-sort the inputs first and then use -sorted with -g when a stable chromosome order matters.
4. If you need to inspect why intervals were altered, rerun with -wo or -wb as a diagnostic pass rather than as the final cleaned output.

Guardrails

• Default subtraction can split one A interval into multiple output fragments; that is expected behavior, not duplication.
• -A removes the whole A feature on qualifying overlap, while -N with -f uses the summed overlap across all B features.
• -wo and -wb change the output layout for overlap inspection; do not treat those outputs like plain trimmed BED intervals.
• -sorted requires truly coordinate-sorted inputs and may need -nonamecheck if naming conventions differ (chr1 vs chr01).
• If you use BAM as input A, add -bed when the downstream consumer expects BED-style output rather than alignment output.