Fig. 2

Overview of the GoPeaks methodology and benchmarking workflow. a Five general steps of the GoPeaks peak calling methodology. Each panel (a1-a5) represents a separate step. (a1) Step indicates the bin width and slide (the width of the bin overlap). (a2) Counting the number of aligned reads per bin. (a3) Example of a Binomial probability test distribution and threshold to retain significantly different peaks. (a4) Filtering out bins with less than 15 counts. (a5) Retained bins within 150 bp are merged and identified as a peak. b Schematic overview of the benchmarking workflow. All CUT&Tag datasets undergo the same pre-processing and are separately analyzed by the peak calling methods. The unprocessed peaks are extracted for sub-analyses. High-confidence peaks were defined by taking the union of statistically significant peaks from all replicates and retaining the peaks present in at least two biological replicates. Generation of the receiver operator characteristic (ROC) and precision-recall (PR) curves require ChIP-seq standards