Bio Chipseq Peak Calling

Approach: Compare treatment BAM signal against input control using MACS3 local Poisson model.

# Call peaks with input control (recommended)
macs3 callpeak -t chip.bam -c input.bam -f BAM -g hs -n sample --outdir peaks/

# For MACS2 (legacy), replace 'macs3' with 'macs2' - syntax is identical

Without Input Control

Goal: Call peaks without a matched input/control sample.

Approach: Use MACS3 with genomic background estimation only (less accurate than with control).

# Not recommended, but possible
macs3 callpeak -t chip.bam -f BAM -g hs -n sample --outdir peaks/

Narrow Peaks (TF, H3K4me3, H3K27ac)

Goal: Call sharp, well-defined peaks typical of transcription factors and active histone marks.

Approach: Use default narrow peak mode with q-value filtering and genome size correction.

macs3 callpeak \
    -t chip.bam \
    -c input.bam \
    -f BAM \
    -g hs \                        # hs=human, mm=mouse, ce=worm, dm=fly
    -n sample_narrow \
    --outdir peaks/ \
    -q 0.05                        # q-value threshold

Broad Peaks (H3K36me3, H3K27me3, H3K9me3)

Goal: Call diffuse, broad enrichment domains typical of repressive or elongation-associated histone marks.

Approach: Enable broad peak mode which links nearby enriched regions into broader domains.

macs3 callpeak \
    -t chip.bam \
    -c input.bam \
    -f BAM \
    -g hs \
    -n sample_broad \
    --outdir peaks/ \
    --broad \                      # Broad peak mode
    --broad-cutoff 0.1             # Broad peak q-value

Paired-End Data

Goal: Call peaks from paired-end sequencing using actual fragment sizes instead of modeled estimates.

Approach: Use BAMPE format so MACS3 calculates fragment size from mate pairs directly.

# MACS3 uses BAMPE format for paired-end
macs3 callpeak \
    -t chip.bam \
    -c input.bam \
    -f BAMPE \                     # Paired-end BAM
    -g hs \
    -n sample_pe \
    --outdir peaks/

Multiple Replicates

Goal: Call peaks from multiple biological replicates pooled together for increased statistical power.

Approach: Provide all replicate BAMs to MACS3, which internally pools reads before peak calling.

# Pool replicates (MACS3 handles internally)
macs3 callpeak \
    -t rep1.bam rep2.bam rep3.bam \
    -c input.bam \
    -f BAM \
    -g hs \
    -n pooled \
    --outdir peaks/

Custom Genome Size

Goal: Call peaks for non-model organisms without a built-in genome size shortcut.

Approach: Provide the effective genome size as a numeric value instead of a species abbreviation.

# For non-model organisms or custom genomes
macs3 callpeak \
    -t chip.bam \
    -c input.bam \
    -f BAM \
    -g 2.7e9 \                     # Effective genome size in bp
    -n sample \
    --outdir peaks/

Common Genome Sizes

Fixed Fragment Size

Goal: Call peaks when fragment size modeling fails or a specific extension size is needed.

Approach: Bypass model building and specify a fixed read extension size manually.

# If modeling fails or for ATAC-seq
macs3 callpeak \
    -t chip.bam \
    -c input.bam \
    -f BAM \
    -g hs \
    --nomodel \                    # Skip model building
    --extsize 200 \                # Fixed extension size
    -n sample \
    --outdir peaks/

Generate Signal Tracks

Goal: Produce normalized signal tracks for genome browser visualization alongside peak calls.

Approach: Enable bedGraph output with signal-per-million-reads normalization, then convert to bigWig.

# Generate bedGraph and bigWig files
macs3 callpeak \
    -t chip.bam \
    -c input.bam \
    -f BAM \
    -g hs \
    -n sample \
    --outdir peaks/ \
    -B \                           # Generate bedGraph
    --SPMR                         # Signal per million reads

# Convert to bigWig (requires UCSC tools)
sort -k1,1 -k2,2n peaks/sample_treat_pileup.bdg > peaks/sample.sorted.bdg
bedGraphToBigWig peaks/sample.sorted.bdg chrom.sizes peaks/sample.bw

Local Lambda for Broad Marks

Goal: Improve broad peak calling by disabling the genome-wide lambda estimate.

Approach: Use --nolambda to rely solely on local background estimation for very broad domains.

# Recommended for very broad marks
macs3 callpeak \
    -t chip.bam \
    -c input.bam \
    -f BAM \
    -g hs \
    --broad \
    --nolambda \                   # Use local lambda only
    -n sample \
    --outdir peaks/

Cutoff Analysis

Goal: Evaluate how different significance thresholds affect the number of called peaks.

Approach: Run MACS3 cutoff analysis mode to generate a table of peak counts at various q-value cutoffs.

# Test different q-value cutoffs
macs3 callpeak \
    -t chip.bam \
    -c input.bam \
    -f BAM \
    -g hs \
    --cutoff-analysis \            # Generate cutoff analysis file
    -n sample \
    --outdir peaks/

Output Files

narrowPeak Format

chr1  100  200  peak_1  100  .  5.2  10.5  8.3  50

Columns: chr, start, end, name, score, strand, signalValue, pValue, qValue, peak

Filter Peaks

Goal: Post-filter called peaks by statistical significance or signal strength.

Approach: Use awk on narrowPeak columns to apply q-value or signal-value cutoffs.

# Filter by q-value
awk '$9 > 2' peaks.narrowPeak > peaks.filtered.narrowPeak  # -log10(q) > 2 means q < 0.01

# Sort by signal strength
sort -k7,7nr peaks.narrowPeak > peaks.sorted.narrowPeak

Key Parameters

Related Skills

• peak-annotation - Annotate peaks to genes
• differential-binding - Compare peaks between conditions
• alignment-files - Prepare BAM files
• chipseq-visualization - Visualize peaks