chromatin-accessibility-bias-correction

chromatin-accessibility-bias-correction

Summary

Correct GC content bias in single-cell or bulk ATAC-seq chromatin accessibility counts before computing deviation scores. This preprocessing step ensures that downstream motif deviation analysis is not confounded by the strong relationship between GC content and chromatin accessibility.

When to use

Apply this skill when you have loaded raw ATAC-seq fragment counts into a SummarizedExperiment object and are preparing to compute motif deviations. GC bias correction must occur early in the workflow, before sample/peak filtering and before computing expectations and deviations, because the bias term becomes part of the statistical model for deviation normalization.

When NOT to use

• Input counts have already been corrected for GC bias by another method or tool.
• The organism or reference genome is not available as a BSgenome package; addGCBias() requires the genome parameter.
• Your analysis goal is clustering or dimensionality reduction without downstream motif deviation analysis; GC bias correction is specific to the deviation workflow and may not benefit purely unsupervised tasks.

Inputs

• SummarizedExperiment object containing ATAC-seq fragment counts (assay slot with count matrix)
• BSgenome object (e.g., BSgenome.Hsapiens.UCSC.hg19) specifying the reference genome

Outputs

• SummarizedExperiment object with updated rowData containing a 'bias' column with GC content bias term for each peak

How to apply

Load the reference genome (e.g., BSgenome.Hsapiens.UCSC.hg19) as a BSgenome object. Call addGCBias() on your counts SummarizedExperiment, passing the genome object as the 'genome' parameter. This function computes the GC content for each peak region and adds a 'bias' column to the rowData slot. The bias values are then used downstream by computeExpectations() and computeDeviations() to generate GC-matched background peak sets and to normalize deviation scores. The rationale is that chromatin accessibility varies systematically with GC content due to sequencing and technical factors; explicitly modeling this bias allows computeDeviations() to compute residuals that reflect true motif-associated variability rather than GC artifacts.