A workflow for simplified analysis of ATAC-cap-seq data in R.

Background: Assay for Transposase-Accessible Chromatin (ATAC)-cap-seq is a high-throughput sequencing method that combines ATAC-seq with targeted nucleic acid enrichment of precipitated DNA fragments. There are increased analytical difficulties arising from working with a set of regions of interest that may be small in number and biologically dependent. Common statistical pipelines for RNA sequencing might be assumed to apply but can give misleading results on ATAC-cap-seq data. A tool is needed to allow a nonspecialist user to quickly and easily summarize data and apply sensible and effective normalization and analysis.
Results: We developed atacR to allow a user to easily analyze their ATAC enrichment experiment. It provides comprehensive summary functions and diagnostic plots for studying enriched tag abundance. Application of between-sample normalization is made straightforward. Functions for normalizing based on user-defined control regions, whole library size, and regions selected from the least variable regions in a dataset are provided. Three methods for detecting differential abundance of tags from enriched methods are provided, including bootstrap t, Bayes factor, and a wrapped version of the standard exact test in the edgeR package. We compared the precision, recall, and F-score of each detection method on resampled datasets at varying replicate, significance threshold, and genes changed and found that the Bayes factor method had the greatest overall detection power, though edgeR was slightly stronger in simulations with lower numbers of genes changed. Conclusions: Our package allows a nonspecialist user to easily and effectively apply methods appropriate to the analysis of ATAC-cap-seq in a reproducible manner. The package is implemented in pure R and is fully interoperable with common workflows in Bioconductor.