07. Peak calling¶

← Previous: BAM filtering and QC · Tutorial contents · Next: Consensus peaks and counting →

Call peaks per biological sample¶

For paired-end ATAC-seq, MACS3 can infer fragments directly from paired alignments with -f BAMPE. Confirm whether narrow or broad peaks fit the biological question and keep the choice consistent across comparable samples.

set -euo pipefail

SAMPLE_ID="sample_01"
FINAL_BAM="results/bam/${SAMPLE_ID}.final.bam"
GENOME_SIZE="CHANGE_ME"
QVALUE=0.05

if [[ "${GENOME_SIZE}" == "CHANGE_ME" ]]; then
  echo "Replace GENOME_SIZE for the selected organism and assembly." >&2
  exit 1
fi

test -s "${FINAL_BAM}"
mkdir -p "results/peaks/${SAMPLE_ID}" logs/peaks

macs3 callpeak \
  -t "${FINAL_BAM}" \
  -f BAMPE \
  -g "${GENOME_SIZE}" \
  -n "${SAMPLE_ID}" \
  --outdir "results/peaks/${SAMPLE_ID}" \
  -q "${QVALUE}" \
  --keep-dup all \
  2>&1 | tee "logs/peaks/${SAMPLE_ID}.macs3.log"

test -s "results/peaks/${SAMPLE_ID}/${SAMPLE_ID}_peaks.narrowPeak"

GENOME_SIZE, the q-value, optional control inputs, duplicate handling, and additional options require study-specific review. The example is not suitable for every assay or organism. It uses --keep-dup all because Chapter 6's example physically removes duplicates before peak calling. If duplicates were retained instead, review the MACS3 duplicate policy rather than copying this option automatically.

Peak gate¶

the caller reports successful completion;
output BED-like files have valid, sorted coordinates on expected contigs;
peak counts, widths, and signal distributions are compared across samples;
fraction of reads in peaks (FRiP) is calculated consistently after peaks are available and interpreted together with depth, TSS enrichment, and complexity;
low-quality or unusual samples are investigated alongside alignment QC;
peak files and the exact BAM inputs are recorded with checksums;
biological replicates remain separate at this stage.

← Previous: BAM filtering and QC · Tutorial contents · Next: Consensus peaks and counting →