Controlling for DNA dosage with whole-genome sequencing improves ATAC-seq peak calling

ATAC-seq is a powerful method for mapping chromatin accessibility and identifying cis-regulatory elements (CREs) across cell types. We reveal that ATAC-seq peak calling is compromised by local biases in DNA dosage, arising from copy number variations but also from DNA replication timing. As part of the FANTOM consortium's efforts, we propose enhancing the model-based analysis of ChIP-seq pipeline by integrating whole-genome sequencing (WGS) data, analogous to input controls in ChIP-seq analyses. This adjustment increases both the number and width of ATAC peaks, with improved proximity to transcription start sites. These ATAC peaks exhibit canonical CRE epigenetic marks and are enriched for trait- and disease-associated genetic variants. Furthermore, the number of peaks called with WGS enhances the correlation with gene expression levels compared to peaks called without WGS control. Collectively, these results demonstrate that integrating WGS as a control significantly enhances the accuracy of ATAC-seq peak calling. Critically, even low-depth WGS data suffice to improve peak calling, making this approach cost effective and scalable. We implemented this method as an open-source Nextflow pipeline for accessibility and reproducibility. By challenging the assumption of uniform genomic visibility, our approach also holds broad implications for other DNA sequencing-based technologies.