macs3-subcommand-chaining

macs3-subcommand-chaining

Summary

Reconstruct the MACS3 callpeak peak-calling pipeline by chaining individual subcommands (filterdup, predictd, pileup, bdgcmp, bdgopt, bdgpeakcall) to progressively transform ChIP-Seq BED/BEDPE reads into narrow peak calls. This skill enables fine-grained control over each step of duplicate filtering, fragment-length estimation, coverage pileup, local-background bias modeling, and statistical scoring.

When to use

When you have aligned ChIP-Seq reads (BED or BEDPE format) and a corresponding control sample, and you need explicit control over peak-calling parameters—including fragment-length prediction, local bias windows (d, slocal=1kb, llocal=10kb), background scaling, and score-cutoff thresholds—rather than using the monolithic macs3 callpeak wrapper.

When NOT to use

• You only have a single replicate and no control sample; local-bias modeling requires control data.
• Your input is already a peak file (BED, narrowPeak, or broadPeak); this skill is for generating peaks from raw reads.
• You are analyzing broad histone marks (e.g., H3K27me3); use macs3 bdgbroadcall instead of bdgpeakcall for the final step.
• Your reads are in formats other than BED/BEDPE (e.g., BAM); convert or use macs3 callpeak wrapper instead.

Inputs

• ChIP-Seq aligned reads in BED format (e.g., CTCF_ChIP_200K.bed.gz)
• Control sample aligned reads in BED format (e.g., CTCF_Control_200K.bed.gz)
• Genome size (human: 2.7e9 bp, or -g hs flag)
• Fragment-length search range (e.g., -m 5 50 for 5–50 bp)

Outputs

• Duplicate-filtered ChIP BED file with final read count
• Duplicate-filtered control BED file with final read count
• Predicted fragment length d (integer)
• ChIP pileup BEDGRAPH track (coverage at fragment-length extension)
• Control d-background BEDGRAPH (d-extended control pileup)
• Control slocal-background BEDGRAPH (1 kb local window)
• Control llocal-background BEDGRAPH (10 kb local window)
• Normalized slocal BEDGRAPH (d/slocal-scaled)
• Normalized llocal BEDGRAPH (d/llocal-scaled)
• Combined maximum-bias BEDGRAPH (max of d, slocal, llocal)
• Scaled local-lambda BEDGRAPH (lambda scaled to ChIP sequencing depth)
• Score BEDGRAPH (q-value or p-value track from bdgcmp)
• Narrow peaks BED file (regions above score cutoff)