Skip to content

ATAC-seq bias correction & differential motif accessibility (Wang)

Bias correction for differential motif-accessibility testing in ATAC-seq — a methods short-talk at EuroBioC2025 by Jiayi Wang on correcting systematic Tn5-insertion-bias and composition artefacts in ATAC-seq before running motif-accessibility differential tests with chromVAR-style pipelines plus limma-voom.

  • Speaker / lead: Jiayi Wang (affiliation TBD from slides)
  • Building on: chromVAR (Greenleaf-lab motif-accessibility deviations) and the limma-voom pipeline
  • Standalone package: no public package name surfaced from the talk title or schedule entry — methods extension; release status TBD
  • Status at EuroBioC2025: methods short-talk — work-in-progress, no public package URL yet identified

Talk at EuroBioC2025

  • Speaker: Jiayi Wang
  • Day / session: Day 2 (Thu Sep 18, 2025) — morning short-talk session, 9:30 AM
  • Talk title: "Bias Correction and Differential Motif Accessibility in ATAC-seq Data"
  • Slides (Zenodo DOI): to verify after publish
  • Video: not recorded
  • Abstract / schedule: eurobioc2025 schedule

What it does

ATAC-seq motif-accessibility analyses (chromVAR and successors) compute per-cell or per-sample motif-deviation scores that summarize accessibility at TF binding sites — but those scores inherit several biases (Tn5 sequence preference, GC-content effects, library-composition shifts) that distort the downstream differential test. The Wang EuroBioC2025 talk presents a correction strategy that adjusts these biases before passing the deviation matrix into a limma-voom differential test, with the goal of getting valid effect sizes and FDR control on motif-accessibility changes between conditions.

How it works

Core idea. chromVAR-style motif-accessibility scoring builds, for each TF motif, a "deviation" — observed accessibility at motif-containing peaks vs an expectation derived from GC-content / accessibility-matched background peaks — and normalizes it into a z-score using a permutation null. The Wang correction sits between this deviation matrix and the downstream differential test, adjusting the deviation scores for residual Tn5 sequence preference, GC content, and library-composition effects before passing them into a limma-voom contrast model. Exact correction model pending speaker's slides / paper.

Inputs / outputs. Input: an ATAC-seq peak × sample (or × cell) count matrix, a motif × peak annotation (e.g., from motifmatchr), and sample metadata for the contrast. Output: a corrected motif × sample deviation matrix plus limma-voom differential-accessibility results (effect sizes, moderated t-statistics, FDR) on motif activity between conditions. Exact output schema pending speaker's slides / paper.

Key innovation. Rather than treating chromVAR's GC-matched background sampling as the whole bias model, the talk argues remaining Tn5 / composition biases still corrupt downstream limma-voom tests and proposes an explicit correction step on the deviation matrix. Specific correction estimator (regression-based residualization vs reweighting vs other) pending speaker's slides / paper.

Parameters worth knowing. Pending speaker's slides / paper. Likely candidates by analogy to the methodological space: background-peak set definition (GC/length matching parameters), choice of bias covariates to regress out (Tn5 k-mer bias track, GC content, peak width, library size), and the limma-voom design / contrast specification.

Canonical example. Pending speaker's slides / paper. No public package URL or worked dataset was surfaced from the EuroBioC2025 schedule entry or talk title at the time of writing; recheck when the Zenodo DOI or preprint is posted.

Where it fits in the corpus

  • AACR 2026: axis = bioinfo / AI methods (epigenomics); valid differential motif accessibility matters for any cis-regulatory cancer-biology dossier
  • EuroBioC 2025: complements footprintR (the Soneson-lab ATAC-seq footprinting talk) — both target ATAC-seq downstream rigor
  • chromVAR upstream: the canonical motif-deviation tool that this work corrects
  • GBCC 2025: no direct counterpart
  • Nextflow Summit 2026: not mentioned

Notes

No standalone package name was confirmed from the EuroBioC2025 schedule entry. The methods are positioned as a correction applied to existing chromVAR + limma-voom workflows rather than a new container, so the eventual release form is more likely to be a small bias-correction R package or a set of functions contributed upstream than a new analysis framework. Listed here as a flagged work-in-progress (external / pre-release) until a package URL surfaces.