ChIP-seq Quality Control

Quality metrics for assessing ChIP-seq experiment success and replicate reproducibility.

FRiP (Fraction of Reads in Peaks)

Measures the proportion of reads falling within called peaks. Higher values indicate stronger enrichment.

Calculate FRiP with bedtools

# Count reads in peaks
reads_in_peaks=$(bedtools intersect -a chip.bam -b peaks.narrowPeak -u | samtools view -c -)
total_reads=$(samtools view -c -F 260 chip.bam)

# Calculate FRiP
frip=$(echo "scale=4; $reads_in_peaks / $total_reads" | bc)
echo "FRiP: $frip"

Calculate FRiP with featureCounts

# Convert peaks to SAF format
awk 'BEGIN{OFS="\t"} {print $4, $1, $2, $3, "."}' peaks.narrowPeak > peaks.saf

# Count reads in peaks
featureCounts -a peaks.saf -F SAF -o peak_counts.txt chip.bam

# FRiP from summary
grep -v "^#" peak_counts.txt.summary

Calculate FRiP with pysam

import pysam
import pybedtools

def calculate_frip(bam_file, peak_file):
    bam = pysam.AlignmentFile(bam_file, 'rb')
    total_reads = bam.count(read_callback=lambda r: not r.is_unmapped and not r.is_secondary)

    peaks = pybedtools.BedTool(peak_file)
    reads_in_peaks = 0
    for peak in peaks:
        reads_in_peaks += bam.count(peak.chrom, peak.start, peak.end)

    frip = reads_in_peaks / total_reads
    return frip

frip = calculate_frip('chip.bam', 'peaks.narrowPeak')
print(f'FRiP: {frip:.4f}')

FRiP Thresholds

Target	Minimum FRiP	Good FRiP
TF (narrow)	0.01	> 0.05
Histone (broad)	0.10	> 0.20
H3K4me3	0.05	> 0.15
H3K27ac	0.05	> 0.10

Cross-Correlation Analysis (NSC/RSC)

Strand cross-correlation analysis measures ChIP enrichment by calculating correlation between read coverage on forward and reverse strands.

Run phantompeakqualtools

# Run SPP cross-correlation analysis
Rscript run_spp.R \
    -c=chip.bam \
    -savp=chip_cc.pdf \
    -out=chip_cc.txt \
    -odir=qc/

# Output columns:
# 1: filename
# 2: numReads
# 3: estFragLen (estimated fragment length)
# 4: corr_estFragLen
# 5: phantomPeak
# 6: corr_phantomPeak
# 7: argmin_corr (minimum strand shift)
# 8: min_corr
# 9: NSC (Normalized Strand Coefficient)
# 10: RSC (Relative Strand Coefficient)
# 11: QualityTag

Interpret NSC and RSC

# Parse results
awk -F'\t' '{
    print "Fragment length:", $3
    print "NSC:", $9
    print "RSC:", $10
    print "Quality:", $11
}' chip_cc.txt

NSC/RSC Thresholds

Metric	Marginal	Acceptable	Ideal
NSC	< 1.05	1.05 - 1.1	> 1.1
RSC	< 0.8	0.8 - 1.0	> 1.0
QualityTag	-2	0	1 or 2

Plot Cross-Correlation in R

library(spp)

chip_data <- read.bam.tags('chip.bam')
binding_characteristics <- get.binding.characteristics(chip_data, srange=c(50, 500), bin=5)

# Cross-correlation plot
pdf('cc_plot.pdf')
plot(binding_characteristics$cross.correlation, type='l',
     xlab='Strand shift', ylab='Cross-correlation')
abline(v=binding_characteristics$peak$x, col='red')
dev.off()

# Extract metrics
print(paste('Fragment length:', binding_characteristics$peak$x))

Library Complexity (NRF, PBC1, PBC2)

Measures of library complexity to detect PCR amplification issues.

Calculate with bedtools

# NRF: Non-Redundant Fraction (unique reads / total reads)
total=$(samtools view -c -F 260 chip.bam)
unique=$(samtools view -F 260 chip.bam | cut -f1-4 | sort -u | wc -l)
nrf=$(echo "scale=4; $unique / $total" | bc)
echo "NRF: $nrf"

# PBC1: PCR Bottleneck Coefficient 1 (M1/Mdistinct)
# M1 = locations with exactly 1 read
# Mdistinct = distinct genomic locations

bedtools bamtobed -i chip.bam | \
    awk '{print $1":"$2"-"$3}' | \
    sort | uniq -c | \
    awk '{
        if($1==1) m1++
        mdist++
    } END {
        print "M1:", m1
        print "Mdistinct:", mdist
        print "PBC1:", m1/mdist
    }'

Library Complexity Thresholds

Metric	Severe	Mild	None
NRF	< 0.5	0.5 - 0.8	> 0.8
PBC1	< 0.5	0.5 - 0.8	> 0.8
PBC2	< 1	1 - 3	> 3

IDR (Irreproducibility Discovery Rate)

Measures consistency between biological replicates by comparing ranked peak lists.

Run IDR Analysis

# Call peaks on each replicate
macs3 callpeak -t rep1.bam -c input.bam -n rep1 -g hs
macs3 callpeak -t rep2.bam -c input.bam -n rep2 -g hs

# Sort by signal value (column 7)
sort -k7,7nr rep1_peaks.narrowPeak > rep1_sorted.narrowPeak
sort -k7,7nr rep2_peaks.narrowPeak > rep2_sorted.narrowPeak

# Run IDR
idr --samples rep1_sorted.narrowPeak rep2_sorted.narrowPeak \
    --input-file-type narrowPeak \
    --rank signal.value \
    --output-file idr_output.txt \
    --plot idr_plot.pdf \
    --log-output-file idr.log

IDR with Pooled Peaks

# Call peaks on pooled data
samtools merge -f pooled.bam rep1.bam rep2.bam
macs3 callpeak -t pooled.bam -c input.bam -n pooled -g hs

# Run IDR with oracle
idr --samples rep1_sorted.narrowPeak rep2_sorted.narrowPeak \
    --peak-list pooled_peaks.narrowPeak \
    --input-file-type narrowPeak \
    --rank signal.value \
    --output-file idr_oracle.txt

Interpret IDR Results

# Count peaks at different IDR thresholds
awk '$5 >= 540' idr_output.txt | wc -l  # IDR < 0.05 (conservative)
awk '$5 >= 415' idr_output.txt | wc -l  # IDR < 0.1 (optimal)

# IDR output columns:
# 1-3: chr, start, end
# 4: name
# 5: scaled IDR (-125 * log2(IDR))
# 6: strand
# 7: signal (from rep1)
# 8: signal (from rep2)
# 9: local IDR
# 10: global IDR

IDR Self-Consistency Check

# Split one sample and check self-consistency
samtools view -s 0.5 chip.bam -b > pseudo_rep1.bam
samtools view -s 2.5 chip.bam -b > pseudo_rep2.bam

# Call peaks on pseudo-replicates
macs3 callpeak -t pseudo_rep1.bam -c input.bam -n pseudo1 -g hs
macs3 callpeak -t pseudo_rep2.bam -c input.bam -n pseudo2 -g hs

# Run IDR
idr --samples pseudo1_peaks.narrowPeak pseudo2_peaks.narrowPeak \
    --input-file-type narrowPeak \
    --output-file self_idr.txt

IDR Quality Guidelines

Comparison	Expected IDR Peaks	Notes
True replicates	> 70% of pooled	Biological concordance
Pseudo-replicates	> 80% of sample	Technical consistency
Rep vs Pooled	~100% of rep peaks	Subset relationship

deepTools QC Metrics

plotFingerprint

# Assess enrichment with fingerprint plot
plotFingerprint \
    -b chip.bam input.bam \
    --labels ChIP Input \
    -o fingerprint.pdf \
    --outRawCounts fingerprint.tab \
    --outQualityMetrics fingerprint_qc.txt

# Good ChIP shows curve shifted right of diagonal
# Input follows diagonal

computeMatrix and plotProfile

# TSS enrichment
computeMatrix reference-point \
    -S chip.bw \
    -R genes.bed \
    --referencePoint TSS \
    -a 3000 -b 3000 \
    -o matrix.gz

plotProfile \
    -m matrix.gz \
    -o tss_enrichment.pdf \
    --perGroup

plotCorrelation

# Sample correlation
multiBamSummary bins \
    -b rep1.bam rep2.bam rep3.bam \
    -o results.npz

plotCorrelation \
    -in results.npz \
    --corMethod spearman \
    --whatToPlot heatmap \
    -o correlation.pdf \
    --outFileCorMatrix correlation.tab

Complete QC Pipeline

#!/bin/bash
sample=$1
input=$2
peaks=$3

echo "=== ChIP-seq QC Report: $sample ===" > qc_report.txt

# FRiP
reads_in_peaks=$(bedtools intersect -a $sample -b $peaks -u | samtools view -c -)
total_reads=$(samtools view -c -F 260 $sample)
frip=$(echo "scale=4; $reads_in_peaks / $total_reads" | bc)
echo "FRiP: $frip" >> qc_report.txt

# Cross-correlation
Rscript run_spp.R -c=$sample -out=cc.txt -odir=.
nsc=$(cut -f9 cc.txt)
rsc=$(cut -f10 cc.txt)
echo "NSC: $nsc" >> qc_report.txt
echo "RSC: $rsc" >> qc_report.txt

# Library complexity
unique=$(samtools view -F 260 $sample | cut -f1-4 | sort -u | wc -l)
nrf=$(echo "scale=4; $unique / $total_reads" | bc)
echo "NRF: $nrf" >> qc_report.txt

# Fingerprint
plotFingerprint -b $sample $input -o fingerprint.pdf --outQualityMetrics fingerprint_qc.txt

cat qc_report.txt

QC Summary Table

Metric	Tool	Ideal Value
FRiP	bedtools/featureCounts	> 0.05 (TF), > 0.1 (histone)
NSC	phantompeakqualtools	> 1.1
RSC	phantompeakqualtools	> 1.0
NRF	samtools/bedtools	> 0.8
PBC1	bedtools	> 0.8
IDR (replicates)	idr	> 70% concordance

Related Skills

• peak-calling - Call peaks before QC analysis
• alignment-files - BAM statistics and filtering
• differential-binding - Compare conditions after QC
• atac-seq/atac-qc - Similar QC for ATAC-seq