1. Duplicate Marking: Sambamba & Picard

Because ChIP-seq yields nanogram quantities of immunoprecipitated DNA, extensive PCR cycles during library preparation can generate significant PCR duplication bias. While natural chromatin fragmentation can produce overlapping reads at hyper-active promoter summits, excessive duplication distorts peak heights. We use sambamba markdup to identify and flag duplicates:

for sample in HD_Rep1 HD_Rep2 Control_Rep1 Control_Rep2 HD_Input Control_Input; do
    echo "Marking PCR duplicates for ${sample}..."
    sambamba markdup -t 6 -r \
        03_filtered/${sample}.q30.bam \
        03_filtered/${sample}.rmdup.bam
    
    samtools index 03_filtered/${sample}.rmdup.bam
done

2. Masking ENCODE Unified Blacklists with bedtools

The ENCODE project has identified anomalous, high-signal regions across the human genome (centromeric repeats, ribosomal DNA, satellite DNA) that non-specifically pull down fragments regardless of antibody specificity. Leaving blacklisted regions unmasked results in false-positive peak calls.

# Download ENCODE v2 GRCh38 Unified Blacklist BED
wget -qO hg38-blacklist.v2.bed.gz "https://github.com/Boyle-Lab/Blacklist/raw/master/lists/hg38-blacklist.v2.bed.gz"
gunzip hg38-blacklist.v2.bed.gz

# Filter out reads overlapping blacklisted coordinates
for sample in HD_Rep1 HD_Rep2 Control_Rep1 Control_Rep2 HD_Input Control_Input; do
    bedtools intersect -v -abam 03_filtered/${sample}.rmdup.bam \
        -b hg38-blacklist.v2.bed > 03_filtered/${sample}.clean.bam
    
    samtools index 03_filtered/${sample}.clean.bam
done

3. Strand Cross-Correlation Metrics: NSC & RSC

Cross-correlation analysis evaluates ChIP enrichment independently of peak calling by assessing the clustering of reads on opposing strands (+ and -) separated by the average sonicated fragment length ($d$). We run phantompeakqualtools (run_spp.R):

• Normalized Strand Cross-correlation (NSC): Ratio of the maximum cross-correlation peak height to the background. Acceptable: NSC ≥ 1.05. High quality: NSC ≥ 1.10.
• Relative Strand Cross-correlation (RSC): Ratio of fragment-length peak height to read-length phantom peak height. Acceptable: RSC ≥ 0.8. High quality: RSC ≥ 1.0.

Rscript /opt/phantompeakqualtools/run_spp.R \
    -c=03_filtered/HD_Rep1.clean.bam \
    -savp=06_qc_reports/HD_Rep1_crosscorr.pdf \
    -out=06_qc_reports/HD_Rep1_crosscorr.txt