differential-chromatin-accessibility-analysis-between-cell-types

differential-chromatin-accessibility-analysis-between-cell-types

Summary

Identify transcription factor motifs showing statistically significant differences in chromatin accessibility bias between distinct cell populations (e.g., GM vs. H1 cell lines) using chromVAR's differential-deviation framework. This skill enables ranking motifs by their variability in accessibility and detecting cell-type-specific regulatory patterns.

When to use

You have pre-processed chromatin accessibility data (ATAC-seq or DNAse-seq) with chromVAR deviations already computed for individual cells or bulk samples across multiple cell types or conditions, and you need to identify which transcription factor motifs exhibit significant differential bias or usage patterns between those populations.

When NOT to use

• Input is raw BAM files or peak counts: must first apply filterSamples, filterPeaks, addGCBias, and computeDeviations to obtain a valid chromVARDeviations object.
• Cell-type or grouping annotation is absent or misaligned in colData: differentialDeviations requires a valid categorical grouping variable in the SummarizedExperiment colData.
• You are performing unsupervised clustering of cells: chromVAR's primary strength for clustering is k-mers + PCA; SnapATAC outperforms chromVAR for clustering tasks according to benchmarks.

Inputs

• chromVARDeviations object with z-score deviations for motifs across samples
• colData annotation specifying cell-type or condition grouping for each sample
• JASPAR motif matches (from matchMotifs step)

Outputs

• variability scores (standard deviation of z-scores) with bootstrap confidence intervals per motif
• differential-deviation test statistics (p-values, effect sizes, bias-corrected deviation estimates) per motif between cell types
• Ranked motif lists sorted by variability or differential significance
• Visualization plots (rank-sorted variability profiles and differential-deviation heatmaps)

How to apply

Start with a pre-computed chromVARDeviations object (dev) derived from filtered counts and matched JASPAR motifs. First, call computeVariability(dev) to generate per-motif standard deviation of z-scores across samples and bootstrap confidence intervals to establish baseline variability. Then call differentialDeviations(dev, grouping_column) where grouping_column specifies the cell-type annotation in colData (e.g., 'Cell_Type' with values 'GM' and 'H1') to perform hypothesis tests of bias-corrected deviations between groups. The function generates p-values and effect sizes per motif; apply a multiple-testing correction threshold (e.g., adjusted p < 0.05) to prioritize significant hits. Export ranked variability scores and differential-deviation test results as structured tables for downstream interpretation and visualization.