Macs3 Peak Calling

# Install with pip or conda
pip install macs3
# or
conda install -c bioconda macs3

# Verify
macs3 --version
# macs3 3.0.2

decisions:
  - id: D1
    param: assayType
    kind: required
    source: user
    ask: "What produced these reads - a transcription-factor ChIP, a histone mark, or open-chromatin (ATAC)?"
    default: null

  - id: D2
    param: controlSample
    kind: required
    source: data
    ask: "Is there a matched input or IgG control for this sample?"
    default: "none - background is estimated locally instead"

  - id: D3
    param: effectiveGenomeSize
    kind: derived
    source: upstream
    ask: "Which genome were the reads aligned to?"
    default: "carried from the alignment reference"

  - id: D4
    param: peakShape
    kind: optional
    source: user
    depends_on: [D1]
    ask: "Are the enriched regions sharp summits or broad domains?"
    default: "narrow for TF and ATAC; broad for spreading histone marks"

  - id: D5
    param: significanceThreshold
    kind: required
    source: user
    ask: "How strong must enrichment be before a region is called a peak?"
    default: "q-value 0.05"

  - id: D6
    param: fragmentModel
    kind: optional
    source: user
    depends_on: [D1]
    ask: "Should fragment length be modelled from the data, or fixed because the assay's signal is not paired summits?"
    default: "model for ChIP; fixed shift/extension for ATAC"

  - id: D7
    param: duplicateHandling
    kind: optional
    source: user
    depends_on: [D1]
    ask: "Are identical read positions PCR duplicates, or real independent insertions?"
    default: "one per position for ChIP; keep all for ATAC"

  - id: D8
    param: signalTracks
    kind: optional
    source: user
    ask: "Emit normalized coverage tracks alongside the peak calls?"
    default: "peaks only"

# Call peaks for TF ChIP-seq (narrow peaks, with input control)
macs3 callpeak \
    -t chip.bam \
    -c input.bam \
    -f BAM \
    -g hs \
    -n sample_tf \
    --outdir peaks/ \
    -q 0.05

# Output: peaks/sample_tf_peaks.narrowPeak
wc -l peaks/sample_tf_peaks.narrowPeak

# Sort and index ChIP and control BAMs (if not already done)
samtools sort -@ 8 chip_raw.bam -o chip.bam
samtools sort -@ 8 input_raw.bam -o input.bam
samtools index chip.bam
samtools index input.bam

# Check read counts
echo "ChIP reads: $(samtools view -c -F 4 chip.bam)"
echo "Input reads: $(samtools view -c -F 4 input.bam)"

# TF ChIP-seq with input control
macs3 callpeak \
    -t chip.bam \
    -c input.bam \
    -f BAM \
    -g hs \
    -n tf_chip \
    --outdir peaks/ \
    -q 0.05 \
    --keep-dup auto

echo "Peaks called: $(wc -l < peaks/tf_chip_peaks.narrowPeak)"
echo "Summit file: peaks/tf_chip_summits.bed"

# Without input control (less recommended)
macs3 callpeak \
    -t chip.bam \
    -f BAM \
    -g hs \
    -n tf_noinput \
    --outdir peaks/ \
    --nolambda

# H3K27me3 broad histone mark
macs3 callpeak \
    -t h3k27me3.bam \
    -c input.bam \
    -f BAM \
    -g hs \
    -n h3k27me3 \
    --outdir peaks/ \
    --broad \
    --broad-cutoff 0.1 \
    -q 0.05

echo "Broad peaks: $(wc -l < peaks/h3k27me3_peaks.broadPeak)"

# H3K4me3 (sharp mark — use narrow peaks)
macs3 callpeak \
    -t h3k4me3.bam \
    -c input.bam \
    -f BAM \
    -g hs \
    -n h3k4me3 \
    --outdir peaks/ \
    -q 0.05

# ATAC-seq with paired-end BAM (recommended)
macs3 callpeak \
    -t atac.bam \
    -f BAMPE \
    -g hs \
    -n atac_sample \
    --outdir peaks/ \
    --nomodel \
    --nolambda \
    -q 0.05 \
    --keep-dup all

echo "ATAC peaks: $(wc -l < peaks/atac_sample_peaks.narrowPeak)"

# Single-end ATAC-seq: shift reads to center on Tn5 cut site
macs3 callpeak \
    -t atac_se.bam \
    -f BAM \
    -g hs \
    -n atac_se \
    --outdir peaks/ \
    --nomodel \
    --shift -100 \
    --extsize 200 \
    --keep-dup all

# Generate bedGraph normalized to million reads (SPMR)
macs3 callpeak \
    -t chip.bam \
    -c input.bam \
    -f BAM \
    -g hs \
    -n chip_track \
    --outdir tracks/ \
    -B \
    --SPMR \
    --keep-dup auto

# Convert bedGraph to bigWig for IGV/UCSC
# Requires bedGraphToBigWig and chrom.sizes
sort -k1,1 -k2,2n tracks/chip_track_treat_pileup.bdg > tracks/chip_sorted.bdg
bedGraphToBigWig tracks/chip_sorted.bdg genome/hg38.chrom.sizes tracks/chip.bw

echo "BigWig track: tracks/chip.bw"

import pandas as pd

# Load narrowPeak file
# Columns: chrom, start, end, name, score, strand, signalValue, pValue, qValue, peak
cols = ["chrom", "start", "end", "name", "score", "strand",
        "signalValue", "pValue", "qValue", "peak"]
peaks = pd.read_csv("peaks/tf_chip_peaks.narrowPeak", sep="\t",
                    header=None, names=cols)

print(f"Total peaks: {len(peaks)}")
print(f"Peaks on chr1: {(peaks['chrom'] == 'chr1').sum()}")
print(f"Median peak width: {(peaks['end'] - peaks['start']).median():.0f} bp")
print(f"Peaks with q-value < 0.01: {(peaks['qValue'] > 2).sum()}")  # -log10(q) > 2

# Filter high-confidence peaks
high_conf = peaks[peaks["qValue"] > 2].copy()  # q < 0.01
high_conf["width"] = high_conf["end"] - high_conf["start"]
print(f"\nHigh-confidence peaks: {len(high_conf)}")
high_conf.to_csv("high_confidence_peaks.bed", sep="\t", index=False, header=False,
                 columns=["chrom", "start", "end", "name", "score", "strand"])