chromatin-accessibility-variability-ranking

chromatin-accessibility-variability-ranking

Summary

Rank transcription factor motifs by their contribution to variability in chromatin accessibility across cell populations, and identify motifs showing statistically significant differential deviation between distinct cell types. This enables discovery of cell-type-specific regulatory patterns driving heterogeneity in chromatin state.

When to use

You have sparse, single-cell or bulk ATAC/DNAse-seq data from multiple cell types or conditions (e.g., GM vs H1 cell lines), pre-filtered and GC-bias-corrected, with motif-to-peak matches already computed. You want to discover which transcription factor motifs contribute most to observed accessibility variation within and between populations, and to test whether motif usage differs significantly by cell type.

When NOT to use

• Input counts have not been filtered for sample depth (min_depth), peak overlap, or GC bias correction — use filterSamples, filterPeaks, and addGCBias first.
• Motifs have not yet been matched to peaks — use matchMotifs and computeDeviations before this skill.
• You seek to cluster cells or perform dimensionality reduction based on chromatin patterns — use SnapATAC or k-mer + PCA approaches instead, as chromVAR has been superseded for that task.

Inputs

• chromVARDeviations object (SummarizedExperiment) with pre-computed deviations from filtered example_counts and matched JASPAR motifs

Outputs

• Ranked motif variability table (standard deviations, bootstrap confidence intervals, hypothesis test p-values)
• Differential deviations test results table (p-values, effect sizes, bias-corrected deviation differences per motif by cell type)
• Variability rank plot (plotVariability output)

How to apply

Load a pre-computed chromVARDeviations object (obtained via computeDeviations on bias-corrected, filtered counts and motif annotations). Call computeVariability() to compute the standard deviation of z-scores across samples for each motif, generate bootstrap confidence intervals by resampling cells/samples, and perform hypothesis tests against a null variability of 1. Rank motifs by their variability score and visualize using plotVariability(). Then call differentialDeviations(dev, "Cell_Type") to test for significant differences in bias-corrected deviations between the two cell groups using the colData cell-type annotation. Export both the ranked variability results and differential-deviation test results (p-values and effect sizes per motif) as structured tables for downstream interpretation.