End-to-end chromatin accessibility analysis covering the nf-core/atacseq pipeline through integration with DNA methylation, histone modifications and gene expression, applied to eight ATAC-seq libraries from developing mouse tissue (Gorkin et al., 2020).
This project processes eight paired-end ATAC-seq libraries from mouse fetal liver and kidney at embryonic days E14.5 and E15.5 (two biological replicates per tissue and timepoint) to identify tissue-specific regulatory elements, infer the transcription factors driving them, and relate accessibility to the other epigenetic layers profiled in parallel by the other course groups. Data are restricted to a reduced mm10 reference containing chromosomes 18 and 19 only (145,658,490 bp).
Data and computing environment provided by the Walter Lab and Müller Lab (ICBB), Saarland University.
Authors: Divyashree Mohan — nf-core/atacseq pipeline, quality control, differential accessibility, motif and footprint analysis, IGV integration, differential methylation vs expression · Ajay Nimbalkar — ChrAccR pipeline and report, methylation–accessibility correlation, DMR overlap analysis
Samples: kidney_14_5_REP1/2, kidney_15_5_REP1/2, liver_14_5_REP1/2, liver_15_5_REP1/2 — 1.7–4.4 M reads each
- Read processing & alignment — nf-core/atacseq v2.1.2 with Singularity: cutadapt trimming, BWA-MEM alignment, Picard MarkDuplicates, ENCODE mm10 blacklist filtering
- Peak calling — MACS2 narrow-peak mode, 1,456–3,415 peaks per replicate merged into a 6,259-peak consensus set
- Quality control — FastQC and MultiQC; FRiP, TSS enrichment, fragment-size periodicity and promoter-overlap metrics
- Signal summarisation — ChrAccR over three region sets (genome-wide tiling, promoters, consensus peaks)
- Filtering & normalisation — low-coverage and chrM removal, then quantile normalisation
- Exploratory analysis — PCA and UMAP per region set, clustered heatmaps of the most variable regions
- Motif activity — chromVAR with 579 JASPAR vertebrate motifs, variability ranking and deviation Z-scores
- Differential accessibility — kidney vs liver on raw counts (padj < 0.01, |log2FC| > 3), volcano plot, nearest-TSS annotation, Wilcoxon test on TSS distances
- TF footprinting — aggregate Tn5 insertion profiles at GATA1, GATA4, GBX2 and HOXD13 motifs from tissue-merged datasets
- Locus-level integration — IGV session combining accessibility, ChromHMM segmentation, WGBS signal and RNA-seq
- Quantitative integration — methylation per chromatin state; methylation vs accessibility across 5,720 exactly matched peaks; DMR overlap with heterochromatin and differential peaks; DMR-to-gene assignment vs expression
- Tissue dominates over developmental time — kidney and liver separate cleanly on consensus peaks (PC1 = 79.69% of variance), while E14.5 and E15.5 samples intermix freely within each tissue (
figures/chraccr/08_pca_consensus_peaks.png) - Differential accessibility is distal — 509 peaks more accessible in liver vs 109 in kidney, but almost none at promoters (7 vs 2); kidney-specific peaks sit significantly farther from the nearest TSS than non-differential peaks (Wilcoxon p = 2.32 × 10⁻¹⁰) (
figures/chraccr/19_volcano_differential_accessibility.png,20_violin_tss_distance.png) - One TF family dominates motif variability — the five most variable motifs are all GATA-family (Gata4 9.18, Gata1 7.80, GATA6 7.26, GATA3 6.72, GATA1::TAL1 6.47), reflecting both hepatocyte identity (GATA4/6) and the erythroid progenitor population of the fetal liver (
figures/chraccr/17_motif_variability.png) - Motif enrichment is not occupancy — footprinting confirms tissue-specific binding at GATA sites, but the kidney-enriched HOXD13 and GBX2 motifs show identical aggregate profiles in both tissues (
figures/chraccr/21_footprint_gata1.png,22_footprint_hoxd13.png) - Methylation and accessibility are inversely related — r = −0.748 (kidney) and −0.523 (liver) across 5,720 matched peaks; restricting to regions differential in both assays tightens this to 535 of 538 overlaps (99.44%) showing the inverse pattern (
figures/integrative/38_scatter_kidney_meth_vs_acc.png,41_dmr_atac_overlap.png) - The methylation–expression relationship is context-dependent — active promoters are unmethylated in both tissues (Pr_A: 0.009 / 0.014) while transcribed gene bodies are the most methylated compartment (Tx_S: 0.898 / 0.812), so gene-averaged methylation correlates only weakly with expression (ρ = −0.10 to −0.12) (
figures/integrative/36_methylation_by_chromatin_state.png)
report/ analysis_report.md — full write-up: methods, 14 figures, interpretation
protocol/ part1.md, part2.md — complete task-by-task workshop notebook
ChrAccR_analysis/ Browsable ChrAccR HTML report (summary, filtering,
normalization, exploratory, differential)
nextflow_run/ samplesheet.csv, config, MACS2 narrowPeak/summits/consensus
integrative/ task2–task5 R scripts, logs and result tables
figures/ 37 figures in pipeline_qc/, chraccr/, integrative/
View the ChrAccR report · Read the analysis report
Alignment files (BAM, bigWig) and MACS2 pileup bedGraphs are excluded — they run to several GB and are regenerable from the pipeline. The 200 bp aggregate count tracks used for IGV can be recreated with exportCountTracks() as documented in protocol/part1.md. Two ChrAccR outputs are also omitted for size: diffObj_tiling.rds (179 MB) and diffTab_1_tiling.tsv (59 MB).
WGBS, ChIP-seq and RNA-seq inputs used in the integrative sections were produced by the wgbs1, wgbs2, chipseq1 and rnaseq1 groups and shared through the course volume.
nf-core/atacseq 2.1.2, Nextflow, Singularity, BWA-MEM, Picard, MACS2, MultiQC, FastQC · R with ChrAccR, chromVAR, DESeq2, GenomicRanges, edgeR, ComplexHeatmap, pheatmap, ggplot2, dplyr · bedtools, deepTools, IGV, ChromHMM · SLURM on the de.NBI cloud
Gorkin, D. U. et al. (2020). The dynamic landscape of open chromatin during mouse embryonic development. Nature 583, 744–751.
Schep, A. N. et al. (2017). chromVAR: inferring transcription-factor-associated accessibility from single-cell epigenomic data. Nature Methods 14, 975–978.