tn5-insertion-bias-correction

Tn5 insertion bias correction

Summary

Corrects systematic sequence preferences in Tn5 transposase insertion patterns within ATAC-seq data to reveal true transcription factor footprints. This bias correction is essential for accurate downstream footprinting analysis, as uncorrected insertion bias can obscure protein-bound depletion patterns.

When to use

Apply this skill when you have aligned ATAC-seq BAM files from Tn5-based chromatin accessibility assays and need to perform footprinting analysis. The skill is triggered when raw insertion signal contains confounding transposase sequence preferences that would mask the depletion patterns (footprints) characteristic of transcription factor binding.

When NOT to use

• Input is non-Tn5 based chromatin data (e.g., ChIP-seq, DNase-seq, or other accessibility assays) — this skill is specific to Tn5 insertion patterns.
• ATAC-seq data is already corrected by another tool — applying bias correction twice may remove legitimate signal.
• You only need peak-level summary statistics and do not require base-pair resolution footprinting — bias correction adds computational cost for minimal benefit in coarse workflows.

Inputs

• aligned ATAC-seq reads (BAM format)
• reference genome sequence (FASTA format)
• peak regions (BED format, optional but recommended)

Outputs

• bias-corrected cutsite signal (bigWig format)
• uncorrected signal track (bigWig format)
• modeled Tn5 bias track (bigWig format)
• expected bias track (bigWig format)
• diagnostic PDF report (ATACorrect.pdf)

How to apply

Load the aligned ATAC-seq BAM file along with the corresponding reference genome FASTA sequence. Run TOBIAS ATACorrect, which models Tn5 insertion bias directly from the input BAM data by learning the sequence-dependent cutting preferences of the transposase. ATACorrect generates four output tracks: uncorrected cutsite signal, modeled bias signal, expected bias signal, and the final bias-corrected signal. The corrected bigWig file removes the learned sequence bias while preserving the footprint signal (insertion depletion around protein-bound sites), producing output suitable for downstream footprint scoring and transcription factor binding analysis.